APPI News — Asia-Pacific reporting for a global readership

Crucial SSD owner reports refund after replacement stock runs out

APPI News Editorial — Wed, 26 Aug 2026 07:06:24 GMT

Windows Central reported on August 18, 2026, that a Crucial X10 Pro 4TB owner completed a Return Materials Authorization (RMA) and was offered the original purchase price after support said no replacement was available. The report relies on the owner's account and a support email translated from German, so it establishes one case rather than a company-wide rule.

Micron announced on December 3, 2025, that it would end shipments under its consumer brand in February 2026 while continuing warranty service and support. The announcement did not promise like-for-like replacements or define how any cash remedy would be valued.

Micron promised service, not a particular remedy

Warranty eligibility and the available remedy are separate questions. A claim can fall within the stated coverage period without giving the owner a contractual right to a new drive if the warranty leaves the remedy to the manufacturer.

The August report shows why that distinction matters during a product wind-down. It does not show that every consumer model or region has run out of replacement units, and Micron has not published a model-by-model inventory list.

The UK terms put two limits on SSD coverage

Micron's UK-facing legal terms cover an SSD bought by the original end customer through the company or an authorized seller for three or five years, depending on the product, or until its published total-bytes-written limit is reached, whichever comes first. The applicable product data sheet sets the limit, and the drive's Self-Monitoring, Analysis and Reporting Technology (SMART) data measures progress toward it.

The same SSD section lets Micron Commercial Products Group choose repair, replacement, store credit or a refund at the lower of the original purchase price and fair market value. A headline warranty period therefore says how long a claim may remain eligible, not which remedy Micron must provide.

Question	What the UK terms say	Evidence to keep
Who may claim?	The original end customer who still owns a product bought through an authorized channel	Dated receipt, seller name and serial number
When does coverage end?	After three or five years, or at the model's total-bytes-written limit, whichever comes first	The product data sheet in force at purchase and a SMART readout
Who chooses the remedy?	Micron Commercial Products Group	The written RMA decision and the clause cited by support
How may a refund be valued?	The lower of original purchase price and fair market value	The invoice and a written calculation from support

The published page leaves a refund ambiguity

After allowing a refund as an SSD warranty remedy, the manufacturer's UK page later says refunds apply only to new products bought directly from its website within 45 days, or 30 days for factory-recertified SSDs. The page does not explain whether that later sentence addresses only ordinary purchase returns or also limits refunds offered under a multi-year warranty.

APPI News could not resolve the two passages from the published text alone. A claimant offered cash should ask support to identify the controlling clause, state the valuation date and formula, and explain in writing whether taxes, shipping or other charges are included.

Local consumer rights remain a separate track

The UK terms say the limited warranty does not affect statutory rights and direct customers to region-specific return routes. The rights available against a seller or manufacturer depend on the country of purchase, the seller's identity and the terms presented when the order was placed.

A manufacturer's warranty should therefore be checked separately from any mandatory rights under local law. The UK page cannot establish the remedy available to a buyer in another country, and APPI News has not verified the rules of every market where the brand's products were sold.

Micron's Client SSD formula does not govern consumer drives

Micron's separate Client SSD warranty, revised January 10, 2017, calculates a credit or refund by multiplying the purchase price by the percentage of useful life remaining. That document names Micron Client SSDs, not the consumer-brand drives, and cannot replace the warranty issued for the product being claimed.

Product line labels matter because different Micron warranty pages use different eligibility periods and refund formulas. The model number on the drive, its packaging and the original sales listing should match the warranty document used in the claim.

Build the evidence file before shipping

Save the exact model and serial number, the dated invoice or order confirmation, the seller's legal name, the product listing, and the data sheet and warranty text available at purchase. If the drive still reports health data, save a SMART screenshot showing its total bytes written and any failure warnings.

Ask support to state the fault decision, available remedies, refund formula, shipping responsibility and remaining coverage after any replacement. Keep the full email thread, the RMA number, packing instructions and shipment tracking rather than relying on a phone summary.

The UK terms require proof of purchase and an RMA number, and instruct owners to back up and delete stored data before sending a drive for repair, replacement, credit or refund. If a failed drive cannot be read or erased, record that fact in the support ticket before shipping it.

Frequently asked questions

Does a five-year SSD warranty guarantee a replacement drive?
No. The UK terms set an eligibility period but leave the choice among repair, replacement, store credit and refund to Micron Commercial Products Group.

Can a refund be lower than the cost of a replacement?
Yes. The UK terms use the lower of the original purchase price and fair market value, while the reported X10 Pro case ended in an offer at the original purchase price even though the owner said a replacement cost more.

Does one reported refund mean Micron changed every consumer SSD warranty?
No. The report documents one X10 Pro 4TB claim, and APPI News could not verify replacement inventory or claim outcomes across other products and regions.

What should be saved before an SSD is shipped?
Keep the receipt, product page, warranty text, model and serial numbers, SMART data, support messages, RMA authorization and tracking record. Back up and erase readable data before return, following the instructions supplied with the RMA.

Apple patches exploited Mac Screen Sharing authentication flaw

APPI News Editorial — Tue, 25 Aug 2026 17:17:57 GMT

The US Cybersecurity and Infrastructure Security Agency added CVE-2026-65400 to its Known Exploited Vulnerabilities catalog on August 18, 2026. Apple had released fixes on August 6 for macOS Sonoma, Sequoia and Tahoe. The flaw can allow an attacker on the network to authenticate to Screen Sharing without valid credentials.

The Netherlands National Cyber Security Centre (NCSC-NL) reported active abuse on multiple systems where port 5900 was reachable from the internet; attackers obtained root access and installed a Monero miner on each system. The advisory does not state how many systems were involved or establish the worldwide scale of exploitation.

Observed attacks reached systems that exposed the Screen Sharing service to the internet. (Illustrative image)

Three macOS branches have documented fixes

Apple identifies macOS Sonoma 14.8.9 as the fixed Sonoma release. The company identifies macOS Sequoia 15.7.9 as the fixed Sequoia release, while its Tahoe notice identifies macOS Tahoe 26.6.1 as the fixed Tahoe release. All three notices carry an August 6, 2026, release date and say improved state management addressed the authentication issue.

Installed macOS branch	Documented fixed release
macOS Sonoma	14.8.9 or later
macOS Sequoia	15.7.9 or later
macOS Tahoe	26.6.1 or later

These are minimum fixed versions for the three named branches, not a complete list of every macOS release Apple supports. APPI News could not verify from the cited security notices whether an older branch received a separate fix. A Mac running an older branch therefore needs a compatibility check through Apple or the organization that manages the device.

Apple documents fixes in Sonoma 14.8.9, Sequoia 15.7.9 and Tahoe 26.6.1.

Observed exploitation matters more than one severity label

The US National Vulnerability Database (NVD) displays a Common Vulnerability Scoring System (CVSS) 3.1 score of 9.8 supplied by the CISA Authorized Data Publisher (CISA-ADP). Its vector describes a network attack of low complexity that needs no privileges or user interaction, but the same page says NVD has not provided its own assessment. NCSC-NL lists 7.1 using a vector that assumes low privileges and narrower impact, so the two published scores are not interchangeable.

The catalog entry and the Netherlands cases establish active exploitation independently of that scoring difference. NVD records the US CISA decision as “active” exploitation with an automatable attack and total technical impact, while the Netherlands report documents root access and miner installation on exposed systems. Those records support prompt remediation without treating either score as a prediction for an individual Mac.

Severity scores describe a vulnerability; they do not measure one Mac's chance of compromise. (Illustrative image)

Check the installed version and run Software Update

Apple says the Apple menu's About This Mac window shows the installed macOS name and version number. Record both before starting the update, then compare the number with the fixed release for that branch. A later compatible release also contains the earlier branch update.

Apple directs Mac owners to Apple menu > System Settings > General > Software Update and recommends making a backup before installing new software. Software Update offers only software compatible with that Mac model. Apple also says installation may restart the Mac several times and may show a progress bar or blank screen.

Open About This Mac and record the macOS branch and full version number.
Open System Settings > General > Software Update, allow the compatibility check to finish and install the offered security update or later release.
Keep the Mac awake and do not close a laptop lid while installation is running.
After the final restart, open About This Mac again and confirm that the version meets or exceeds the fixed release in the table.

An organization-managed Mac may receive updates on an administrator's schedule or show controls that the user cannot change. Record the displayed version and contact the organization's IT or security team instead of removing management profiles. A message that no compatible update is available needs follow-up if the installed branch is below the documented fixed version.

Version verification after the restart confirms whether the fixed release was installed. (Illustrative image)

Turn off Screen Sharing if the update must wait

Apple's Mac User Guide places the control at Apple menu > System Settings > General > Sharing and says other computers on the network cannot connect to the Mac when Screen Sharing is off. Turning the service off removes that connection path while an update is pending. It does not patch CVE-2026-65400 or show whether a connection occurred earlier.

The same guide says Screen Sharing and Remote Management cannot be enabled at the same time. Apple's CVE notices name Screen Sharing; they do not state that CVE-2026-65400 affects Remote Management. Managed-device owners should ask their administrator which remote service is enabled instead of assuming the two controls have the same vulnerability.

Disabling an unused remote service can reduce exposure while the system update is pending. (Illustrative image)

An update does not clear evidence of an earlier intrusion

The NCSC-NL cases show that successful exploitation can lead to root access and software installation. A Mac that exposed port 5900 before it was fixed, or that now shows unexplained accounts, files or sustained processor use, needs more than a version check. Disconnecting that Mac from the network and handing it to an organization's incident-response team preserves the option for a controlled review.

A US CISA bulletin says its Binding Operational Directive 26-04 requires US federal agencies to check for pre-patch compromise in designated scenarios. That directive governs US federal agencies, not Mac owners worldwide. The narrow lesson for other organizations is that installing a fix changes the current software but does not determine what happened before installation.

Frequently asked questions

Which Macs need the documented update?
A Mac on Sonoma below 14.8.9, Sequoia below 15.7.9 or Tahoe below 26.6.1 is below the fixed release documented for its branch. Install the compatible update offered by Software Update and confirm the version after restart.

Does a Mac still need the update when Screen Sharing is off?
Turning Screen Sharing off closes that service's network connection path, but it does not alter the vulnerable system component. Installing the compatible fixed release prevents the setting from becoming an unpatched entry point if the service is enabled later.

Does installing the update prove the Mac was not compromised?
No. The update fixes the authentication issue from that point forward, while an earlier intrusion can leave accounts, files or running software behind. A previously internet-exposed or suspicious managed Mac needs review under the organization's incident process.

Why do published CVSS scores differ?
The CISA-ADP and Netherlands NCSC records use different assumptions about required privileges and impact. Cite the provider with the score, and use the confirmed exploitation reports when setting remediation priority.

Sleep apnea diagnosis requires testing beyond a snoring checklist

APPI News Editorial — Tue, 25 Aug 2026 12:00:00 GMT

A 2025 clinical guideline from the US Departments of Veterans Affairs and Defense strongly recommends polysomnography or home sleep apnea testing for clinically suspected obstructive sleep apnea (OSA). The same guideline treats the STOP questionnaire as a screening tool that identifies people who need further evaluation, not as a diagnostic test.

Snoring is one clue, not the dividing line between primary snoring and OSA. The US National Heart, Lung, and Blood Institute lists breathing that starts and stops, frequent loud snoring, gasping, daytime sleepiness, morning headaches and repeated nighttime urination among possible sleep apnea symptoms.

Awake breathing trouble and dangerous sleepiness need immediate action

A witnessed pause during sleep belongs in a clinical assessment, but it is different from a person suddenly stopping breathing while awake. MedlinePlus, a service of the US National Library of Medicine, calls for the local emergency number when someone collapses, cannot breathe or is unresponsive; a routine sleep appointment should not delay that response.

The American Academy of Sleep Medicine says drowsiness reduces attention, delays reactions and impairs decisions, and it advises a sleepy driver to leave the road at a safe location. Repeated difficulty staying awake while driving or doing safety-sensitive work calls for the activity to stop and for prompt medical assessment.

The symptom pattern matters more than snoring volume

The symptom pattern can include breathing that repeatedly stops and restarts, loud snoring, gasping, daytime sleepiness, fatigue, poor concentration, a dry mouth, morning headache or frequent nighttime urination. A bed partner may notice the nighttime events before the person affected does.

The American Academy of Sleep Medicine's 2017 adult guideline defines increased risk of moderate to severe OSA as excessive daytime sleepiness plus at least two of three findings: habitual loud snoring, witnessed apnea or gasping or choking, and diagnosed hypertension. That combination helps clinicians select uncomplicated adults for testing; it does not diagnose OSA at home.

A questionnaire can flag risk but cannot diagnose OSA

STOP and STOP-Bang ask about features associated with OSA and can structure an initial assessment. The 2017 American Academy of Sleep Medicine guideline strongly recommends against using clinical tools, questionnaires or prediction algorithms to diagnose adult OSA without polysomnography or home sleep apnea testing.

A high score therefore means that further evaluation may be warranted, not that treatment should begin automatically. A low score also cannot explain every cause of daytime sleepiness, morning headache or disrupted sleep.

Home and laboratory tests serve different patients

The US National Heart, Lung, and Blood Institute says laboratory polysomnography records brain waves and monitors heart rate, breathing and blood oxygen during sleep. The 2017 American Academy of Sleep Medicine guideline calls polysomnography the standard diagnostic test when a comprehensive sleep evaluation raises concern for adult OSA.

The American Academy of Sleep Medicine limits home sleep apnea testing to uncomplicated adults with signs and symptoms that indicate increased risk, and says it should not screen an asymptomatic population. Under that US position, a clinician should order the test and a sleep physician should review the raw data rather than relying only on automatic scoring.

The academy recommends laboratory polysomnography after one negative, inconclusive or technically inadequate home test. It also favors laboratory testing over a home test for people with significant cardiorespiratory disease, possible respiratory muscle weakness, awake or suspected sleep-related hypoventilation, chronic opioid use, a history of stroke or severe insomnia.

Treatment starts after objective diagnosis

The American Academy of Sleep Medicine's 2019 guideline strongly recommends positive airway pressure (PAP) for adults with diagnosed OSA and excessive sleepiness. It says PAP should follow objective testing and should include follow-up that checks effectiveness, use and problems with the equipment.

Continuous positive airway pressure (CPAP) delivers a set pressure, while auto-adjusting positive airway pressure (APAP) changes pressure during sleep. The 2019 guideline strongly recommends either CPAP or APAP for ongoing treatment in adults, but the device, mask and settings require an individualized plan.

A 2015 joint guideline from the American Academy of Sleep Medicine and American Academy of Dental Sleep Medicine says sleep physicians should consider oral appliances for adults who cannot tolerate CPAP or prefer an alternative. When an appliance is prescribed, the guideline favors a custom, titratable device fitted by a qualified dentist over a noncustom device.

The same guideline calls for dental oversight to detect tooth or bite changes and for follow-up sleep testing to confirm effectiveness. An over-the-counter mouthpiece cannot establish that OSA is controlled, even if snoring becomes quieter.

Weight management applies when overweight or obesity is present

The American Thoracic Society's 2018 guideline strongly recommends a comprehensive lifestyle intervention for adults with OSA who are overweight or obese, combining a reduced-calorie diet, physical activity and behavioral guidance. The reviewed interventions were associated with improvements in apnea severity and quality of life, but the studies had methodological limits.

The recommendation does not mean that every adult with OSA is overweight or that weight loss guarantees the condition will disappear. The US National Heart, Lung, and Blood Institute says follow-up testing may be needed after substantial weight gain or loss, so another prescribed treatment should not change without reassessment.

Long-term risks are not an individual forecast

The US National Heart, Lung, and Blood Institute says untreated sleep apnea may raise the risk of cardiovascular disease, stroke, type 2 diabetes and other conditions. These are population-level associations and do not predict which complication, if any, one person will develop.

Effective treatment can target breathing events, oxygen disruption, sleepiness and sleep-related quality of life. The cited guidelines do not support promising that one treatment will prevent every heart attack, stroke or diabetes diagnosis.

Children and pregnancy follow separate pathways

The American Academy of Pediatrics' 2012 childhood guideline recommends polysomnography or specialist evaluation when a child regularly snores and has signs or symptoms of OSA. Children may show hyperactivity, bedwetting or school attention problems, so the adult symptom threshold and home-testing rules should not be copied into pediatric care.

A July 2026 American College of Chest Physicians guideline issued nine conditional recommendations for OSA in pregnancy, including risk screening, objective testing and postpartum reassessment for some patients. The panel rated the certainty as low, so pregnancy requires an obstetric and sleep-medicine plan rather than the routine adult pathway described above.

The cited US guidelines do not create a worldwide testing or treatment policy. APPI News could not verify one global rule for sleep-test eligibility, device prescribing, regulatory status or coverage; local health systems set those arrangements.

Frequently asked questions

Does loud snoring mean OSA?
No. Loud snoring can raise suspicion, especially with witnessed breathing pauses, gasping or daytime sleepiness, but an objective sleep test is needed for diagnosis.

Can a questionnaire confirm OSA?
No. STOP, STOP-Bang and similar tools can identify risk and support referral, but the guidelines cited here do not accept them as stand-alone diagnostic tests.

Does a negative home test rule OSA out?
Not always. A negative, inconclusive or technically inadequate home test can require laboratory polysomnography or another test when clinical suspicion remains.

Is a noncustom anti-snoring mouthpiece equivalent to a prescribed appliance?
No. The joint sleep and dental guideline favors a custom, adjustable appliance with dental oversight and follow-up sleep testing when oral appliance therapy is chosen.

Is PAP the only treatment?
No. PAP has strong evidence for adults with OSA and excessive sleepiness, while a prescribed oral appliance or weight management for people with overweight or obesity may fit different cases.

POCO F9 series sales in Taiwan start September 4 at noon

APPI News Editorial — Tue, 25 Aug 2026 05:06:23 GMT

Xiaomi Taiwan opened POCO F9 series reservations on August 24 ahead of first-sale coupon use from noon on September 4, 2026 (GMT+8). The company has previewed series-level chip, display and charging claims, but Taiwan prices and complete local configurations remain unverified.

The sale covered here is specific to Taiwan. Prices, colors, memory combinations, warranty terms and accessories announced in other markets do not establish what Xiaomi will sell there.

A launch preview establishes only part of the information needed for a phone comparison.

Reservations close one minute before coupon use begins

Xiaomi Taiwan's campaign terms say reservations run from 4:00 p.m. on August 24 through 11:59 a.m. on September 4, 2026 (GMT+8). The first-sale coupon becomes usable one minute later, at noon.

The terms set the coupon at NT$1,000 and make it valid through 11:59 p.m. on September 21. Using Taiwan's August 24 closing rate of NT$31.854 per US$1, the coupon is worth about US$31. Xiaomi says it can be used once, cannot be combined with another coupon and becomes invalid if the phone is returned.

The Taiwan coupon has separate reservation and redemption periods.

The specifications remain a series-level preview

Xiaomi Taiwan identifies the F9 line with Snapdragon 8 Elite Gen 5-family chipsets. The page says the 8,050mAh battery and independent subwoofer are exclusive to the F9 Ultra, while the luminous rear panel belongs only to the Aurora Green F9 Pro.

Xiaomi's English-language preview says the Ultra uses Snapdragon 8 Elite Gen 5 and the Pro uses a V-series version of the chipset family. It also advertises 100W wired and 50W wireless charging, an AMOLED display rated up to 185Hz and Bose-branded symmetrical stereo speakers.

The official preview describes 185Hz as a maximum refresh rate and limits its 185fps claim to selected games under internal test conditions. Xiaomi also directs customers to local sellers for available colors, storage and RAM variants. Those qualifiers prevent a series-level preview from proving that every Taiwan configuration has every advertised feature.

Chip, memory and storage combinations still need model-by-model confirmation.

Taiwan prices remain missing

As of August 25, Xiaomi Taiwan's launch page carried reservation offers but no F9 list price that APPI News could verify. The campaign terms also omitted model-by-model prices, RAM, storage and camera details. No online estimate is used here.

Xiaomi Taiwan's current POCO list starts the X8 Pro at NT$10,999, the F8 Pro at NT$16,999 and the F8 Ultra at NT$21,999. At the August 24 closing exchange rate, those prices are about US$345, US$534 and US$691. They are local comparison points, not forecasts for the F9 series.

The F8 Pro's Taiwan page lists a Snapdragon 8 Elite chipset, a 6,210mAh battery, 100W wired charging and a 120Hz AMOLED display. A useful F9 comparison therefore requires the final price and exact configuration, not a processor name or charging figure in isolation.

Battery, charging, software support and price affect different parts of the ownership period.

A purchase comparison starts after model pages go live

The first comparison is the final price of each RAM and storage configuration after any applicable coupon. Only then can the F9 Pro and Ultra be measured against current POCO models or phones from another brand.

If Taiwan receives both 12GB and 16GB variants, APPI News's RAM guide treats the choice as a question of app reloads and documented on-device features, not a general speed rating. Storage capacity and the price difference can then be assessed separately.

Software support also needs a stated term. APPI News's phone-update guide separates support status and required-app compatibility from other hardware specifications. The final Taiwan listing needs to establish operating-system and security support, charger contents, local warranty and repair coverage.

Until those pages appear, a reservation establishes coupon eligibility but not the value of either phone. The comparison can be completed once Xiaomi publishes Taiwan prices, exact memory and storage options, camera specifications, charger contents and support terms.

The final choice depends on the exact model, price and support policy offered in the buyer's market.

Frequently asked questions

When do POCO F9 sales start in Taiwan?
Xiaomi Taiwan's terms close reservations at 11:59 a.m. on September 4, 2026 (GMT+8), and make the first-sale coupon valid from noon. Noon is therefore the documented first-sale time used in this report.

Has Xiaomi announced Taiwan prices?
APPI News could not verify a Taiwan list price for the F9 Pro or F9 Ultra on August 25. The official pages carried promotion and preview information, while online price estimates lacked support from a Taiwan product listing.

What differences between the F9 Pro and Ultra are confirmed?
Xiaomi reserves the 8,050mAh battery and independent subwoofer for the Ultra. The Aurora Green Pro gets the luminous rear panel, but complete Taiwan memory, storage, camera and color combinations remained unlisted.

Can current POCO prices predict the F9 price?
No. The X8 Pro, F8 Pro and F8 Ultra provide local price benchmarks, but they do not establish where either F9 model will be priced or which configurations Taiwan will receive.

FHIR data exchange leaves six checks for medical AI deployment

APPI News Editorial — Tue, 25 Aug 2026 00:00:00 GMT

Taiwan's Ministry of Health and Welfare launched a medical informatics certification laboratory on August 18, 2026. The ministry, Taiwan's central health and welfare policy agency, plans certification tracks for software, hardware running Fast Healthcare Interoperability Resources (FHIR) Box and solution providers. FHIR Box is Taiwan's interoperability and application layer above existing hospital systems.

The planned tests do not show whether connected records are suitable for an artificial intelligence model or whether a model will perform safely at another hospital. The distinction matters wherever a health system uses an interoperability layer to connect medical AI with electronic records. The six checks below separate technical conformance from the clinical, operational and governance evidence needed for deployment.

FHIR solves exchange, not fitness for AI

HL7 International defines FHIR as a standard for exchanging health information electronically and describes Resources as its basic building blocks. Profiles can constrain fields, cardinality, terminology bindings, data types and extensions for a particular implementation. Those rules describe the shape and meaning of an exchange, not the accuracy of the source record.

Two hospitals can use different FHIR releases, implementation guides, profiles, extensions and code systems while both describe their systems as FHIR-based. Each deployment therefore needs a mapping record that identifies required fields, local codes, units, time stamps and the handling of missing or invalid values. Model outputs written back to a record also need provenance, model version, time stamps, user actions and an audit trail.

HL7's SMART App Launch guide defines OAuth 2.0-based patterns for applications to authorize, authenticate and connect to FHIR systems, with scopes that limit delegated access. That access layer does not determine whether a hospital has granted the right people the right permissions or whether the data and model are clinically reliable.

Conformant records can still contain poor data

A laboratory value can fit a valid FHIR resource while carrying the wrong unit, a shifted time stamp or a local code that was mapped incorrectly. Missing values and differences in how hospitals label diagnoses or outcomes can also alter the population represented in a dataset. An interface can transmit each problem without detecting it.

A 2023 systematic review led by Abigail Lewis included 103 papers and found that completeness was the most common data-quality dimension, followed by correctness, concordance, plausibility and currency. The authors also identified conformance and bias as additional dimensions. They concluded that no standard approach for assessing electronic health record data quality existed across the literature reviewed.

External validation must match intended use

A model evaluated at its development hospital may encounter different patients, devices, prevalence and workflows elsewhere. Overall performance can hide weaker results in groups that were uncommon in the test data. Local acceptance testing should therefore use the deployed model version, intended population and actual data path.

The International Medical Device Regulators Forum's 2025 principles call for representative datasets, independent training and test sets, external validation proportionate to risk, assessment of the human-AI team and monitoring after deployment. The test record should state the intended use, reference standard, sites, sample selection, subgroup results and the role of clinical staff in reviewing the output. FHIR conformance and external validation belong in the same procurement record, but one cannot substitute for the other.

Privacy and responsibility sit outside the wire format

The World Health Organization's 2021 health AI guidance sets six consensus principles, places ethics and human rights at the center of design, deployment and use, and calls for public and private actors to remain accountable to health workers and affected communities. A compliant data format does not assign that accountability.

A deployment record should identify the purpose of each data use, the minimum fields required, who can access them, how long records and logs are retained, and who responds to incidents. Contracts should assign responsibility for mapping errors, unauthorized access, model failures, security patches and service termination. The legal basis and required patient information must be checked under the laws that apply in each country and use case.

Explainability is only one part of transparency

Joint 2024 principles from the US Food and Drug Administration, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency treat explainability as one part of transparency for machine learning-enabled medical devices. They call for information on intended use, target populations, workflow, performance, risks, data gaps, local testing, ongoing monitoring and product changes. The amount and form of information should reflect the device's risks and its intended users.

A 2023 systematic review of explainable AI in health care found only six eligible studies among 882 records and reported no comprehensive, agreed framework or standardized approach for evaluating explanation effectiveness. Feature importance or another single explanation method therefore cannot serve as a general safety test. Explanations must be assessed alongside clinical performance, workflow controls and monitoring.

Six checks keep deployment claims in scope

A procurement or governance record can group the evidence into six areas. Each item should name the exact software version, site and intended use covered by the evidence so that a broad platform claim does not stand in for a product-specific test.

Interoperability: Record the FHIR release, profiles, terminology bindings, implementation guide, validation results and local mapping rules.
Data quality: Measure missing values, invalid codes, units, time stamps, outliers, label quality and representation of the intended population.
External validation: Report performance across sites, devices and patient groups, including uncertainty and groups with small samples.
Clinical workflow: Define inputs, outputs, human review, overrides, escalation, downtime procedures and alert management.
Ongoing monitoring: Keep data and model versions, drift measures, subgroup performance, security incidents, manual overrides and update records.
Accountability: Assign hospitals, platform operators, integrators and AI suppliers responsibility for errors, failures, incidents and contract exit.

Comparative evidence remains limited

The two systematic reviews cited here address electronic record quality and explanation effectiveness, not the clinical outcomes or cost of FHIR-based medical AI deployments across hospitals. The three-regulator principles concern transparency for machine learning-enabled medical devices and do not determine whether a product is authorized in a particular country.

As of August 26, 2026, APPI News could not find a published count of completed FHIR Box deployments or a public roster of products certified through Taiwan's planned tracks. APPI News also could not find international comparative data showing that a FHIR-based integration layer improves cross-hospital clinical outcomes or lowers deployment costs. Those gaps do not show there is no benefit; they limit how far interoperability evidence can support a clinical or economic claim.

Hong Kong researchers test 10-item screen for older adults’ social risk

APPI News Editorial — Tue, 25 Aug 2026 00:00:00 GMT

Hong Kong Polytechnic University (PolyU) on August 20, 2026, presented a 10-item screen intended to identify social vulnerability among older adults in community and primary-care settings. The test covers resources, participation, connections, relationships and self-management, but its validation data come from 234 adults in Hong Kong.

A paper published online on June 3, 2026, reported acceptable model fit and strong internal consistency for the shorter scale. Its authors called for independent, longitudinal and cross-cultural validation before clinical use.

Urgent health changes need an emergency route

A social-risk screen cannot assess an acute medical problem. The World Health Organization (WHO) says chest pain or discomfort, difficulty breathing, sudden weakness on one side, new trouble speaking or seeing, loss of balance, a severe unexplained headache, fainting or unconsciousness require immediate medical care.

Local emergency numbers and access routes differ by country. The SF-10 research concerns scheduled assessment of social vulnerability, so its score cannot replace urgent assessment or a clinical evaluation.

Social frailty is broader than loneliness

A 2017 scoping review of 42 studies described social frailty as a continuum in which a person is at risk of losing, or has lost, resources needed to meet basic social needs. The authors’ model also included social activities, social behavior and self-management.

WHO defines social isolation as the objective state of having too few roles, relationships and interactions, while loneliness is distress caused by a gap between the connections a person has and those they want or need. Living alone can contribute to isolation, but it does not by itself establish loneliness or social frailty.

Social connection also has structural, functional and qualitative dimensions. A screen can therefore ask whether support exists and whether relationships remain usable, rather than treating household size as a complete measure.

The shorter scale performed better in one Hong Kong sample

The researchers used a three-phase scale-development process and administered an initial 15-item Social Frailty Index to 234 community-dwelling adults aged 60 or older. PolyU said two rounds of co-design workshops involved 40 older adults, family caregivers, social workers, occupational therapists, nurses and doctors.

The 15-item version failed the study’s model-fit test, while the 10-item version achieved acceptable fit and strong internal consistency. Those findings show how the items behaved within the study sample; they do not establish prediction of future health, service use or response to an intervention.

SF-10 retained items covering general resources, social participation, social connections, interpersonal relationships and self-management. The authors said the total score could identify overall social vulnerability and that domain-level results might guide tailored assessment and referral, while warning that subscale interpretation needs more validation.

The accessible abstract does not report a threshold that services can apply across countries. APPI News could not verify a validated cross-country cutoff as of August 26, 2026.

PolyU said social workers in nongovernmental organizations and community clinics had adopted the tool. The announcement did not name those organizations or provide adoption totals, referral completion rates or outcomes, and APPI News could not find those data in a published source.

ICOPE places screening inside a broader care pathway

WHO’s 2025 Integrated Care for Older People (ICOPE) handbook describes locally adaptable primary and community care pathways that detect declines in intrinsic capacity, identify social care and support needs, and develop a personalized care plan. SF-10 is narrower: it focuses on social vulnerability rather than the wider assessment and planning process.

WHO’s basic ICOPE assessment covers cognition, mobility, vitality, vision, hearing and psychological capacity before a fuller assessment. Neither an ICOPE signal nor an SF-10 score establishes a diagnosis, explains the cause of a change or decides access to a country’s long-term care benefits.

The distinction matters because a screening result and an administrative eligibility decision answer different questions. Health and social care entry points, benefit rules and the professionals responsible for follow-up vary by jurisdiction.

A completed referral matters more than a sent form

The ICOPE pathway continues from basic screening through in-depth assessment, a personalized care plan, implementation and monitoring. It offers a transferable principle for SF-10 programs: services need to record whether a person reached the next step, not only whether a questionnaire was completed.

For SF-10, a practical handoff can be divided into five checkpoints. This sequence is an APPI News synthesis of the cited SF-10 and ICOPE material, not a validated SF-10 protocol or a universal timetable.

Review the result: Record the domain that raised concern and the conclusions that the screen cannot support.
Assess the need: Examine health, communication, transport, financial and relationship factors that may explain the reported difficulty.
Agree on a goal: Record the older person’s priorities and consent before sharing information or arranging support.
Confirm the handoff: Name the receiving local service and document whether it accepted the referral and made contact.
Monitor access: Check whether cost, waiting time, language, disability access or transport prevented use, then reassess when circumstances change.

A sent referral shows that a request left one service; it does not show that another service accepted it or that support began. Without those later records, a program cannot distinguish screening reach from actual access.

Cross-cultural use and service capacity remain unresolved

The authors described SF-10 as culturally grounded for older Chinese adults in Hong Kong and explicitly called for cross-cultural validation. That evidence does not show that translated wording, response patterns or thresholds perform the same way in another population.

Programs also need consent and access controls because the domains concern relationships, resources and support networks. The accessible study abstract does not report whether digital, paper or interviewer-led administration changes participation, privacy or refusal rates.

Service capacity is a separate constraint. Screening can expose transport, financial or social-support needs, but the cited sources do not show that a suitable local service is available, affordable or able to accept each referral.

The evidence applies to older adults, not every population

The study enrolled adults aged 60 or older and provides no validation evidence for children or pregnant people. A child or pregnant person with a change in health or function needs clinician-led assessment rather than an older-adult social-frailty score.

Chronic illness, disability, sensory loss and regular medicines can affect social participation and the appropriate follow-up. An SF-10 result cannot determine the cause of those changes or serve as a basis for starting, stopping or changing medical treatment.

Initial validation supports a limited claim

SF-10 has initial psychometric evidence as a brief social-vulnerability screen for community-dwelling older adults in Hong Kong. The study does not establish a universal cutoff, predict outcomes over time or show that completing the screen reduces disability, hospital admission or mortality.

The next evidence gap is operational as well as statistical: independent studies need to test the scale in other populations, while services need to report whether flagged needs lead to accepted referrals and sustained support. Until then, the tool is an assessment entry point rather than proof of an effective care pathway.

Cloudflare designs deployment docs and tools for AI coding agents

APPI News Editorial — Tue, 25 Aug 2026 00:00:00 GMT

Cloudflare's agent documentation, updated June 24, 2026, says every documentation page can be returned as Markdown and lists skills, llms.txt indexes and managed Model Context Protocol (MCP) servers. These resources let a compatible agent retrieve product instructions and, with configured permissions, call Cloudflare application programming interfaces (APIs) or inspect logs. They provide a plausible route into an agent's working context, but they do not establish how often Cloudflare is recommended.

Cloud-platform advice during vibe coding is shaped by more than the model's memory of a brand. A coding assistant may retrieve current documentation, use an installed command-line interface (CLI) or MCP server, and filter candidates against the framework, data, cost and account constraints in its context. APPI News could not find a public dataset that measures recommendation share across coding assistants under controlled conditions.

Vibe coding still ends in an operations chain

Vibe coding uses natural-language instructions to direct an AI system that creates files, installs packages, changes code and runs tests. A generated page is only an early deliverable. A public service still needs a reproducible build, a runtime, secrets, Domain Name System (DNS) records, Transport Layer Security (TLS), monitoring and a rollback path.

A common route puts the code in a Git repository and lets a cloud platform build each change. Cloudflare says a Pages project linked to GitHub or GitLab deploys on branch pushes and creates preview URLs for branches and pull requests. A short, documented path can reduce the number of decisions an agent must encode, although the deployment still needs human-defined acceptance checks.

Cloudflare has made its workflow legible to agents

Cloudflare Workers can deploy Worker code and static assets in one operation, with Cloudflare handling the assets' caching and delivery. That arrangement can remove a handoff between a static host and a separate runtime. It does not prove that the runtime supports every framework feature or that the resulting service will be cheaper.

Cloudflare also exposes several ways for an agent to obtain instructions. Its documentation offers page-level Markdown, sitewide and product-specific llms.txt indexes, task-specific skills, an OpenAPI specification and managed MCP servers that can search documentation or call Cloudflare services. These interfaces can turn a prose suggestion into a sequence that a compatible tool can attempt and verify.

Discoverability, operability and suitability remain separate tests. A platform can be easy for an agent to find but unavailable through the tools in a particular session. It can also be executable yet fail the project's runtime, data-location, cost or governance requirements.

Four platforms expose different shortest paths

Cloudflare, Vercel, Netlify and Firebase all document routes from source code to a live site. Their differences become clearer when the same project is compared by framework behavior, preview workflow, managed data services, network integration, portability and measured cost. Vendor documentation establishes supported paths, not a neutral ranking.

Platform	Documented deployment path	What a project test must resolve
Cloudflare Workers and Pages	Workers can ship code with static assets, while Pages can build GitHub or GitLab branches and create previews.	Framework and runtime compatibility, binding behavior, observability, resource limits and the work required to move away.
Vercel	Vercel's machine-readable index directs agents to Git or CLI deployment, its REST API, an MCP server and isolated sandboxes for agent-generated code.	Framework-specific behavior, preview collaboration, usage-based costs and portability of platform services.
Netlify	Netlify says it detects and configures projects for a documented range of frameworks. It supports Git continuous deployment and says file-changing runs by its dashboard AI agents produce a Deploy Preview.	Adapter behavior, build settings, preview controls, resource accounting and any manual steps in recovery.
Firebase App Hosting	Firebase documents built-in support for Next.js and Angular, GitHub-triggered rollouts, Cloud Build builds, Cloud Run serving and Cloud CDN caching.	Fit with Firebase and Google Cloud services, supported regions, account and billing setup, runtime behavior and exit work.

The shortest first deployment is not necessarily the best production path. Teams need to run the same minimal application on each serious candidate, record every manual intervention and test rollback, logs, secrets and failure recovery. The site's workload-based AI model selection method applies the same principle: define pass-or-fail requirements before comparing provider claims.

A cross-border review also needs the current plan price, invoice currency, tax treatment, service region and support terms for the actual account. Projects handling personal or confidential data must identify storage and backup locations, subprocessors, administrator access, log retention, deletion procedures and export options. The applicable requirements depend on the countries, contracts and data involved.

An AI recommendation is not a single ranking

A coding assistant can reach a platform name through several routes. Its trained model may associate the task with public examples, a retrieval system may fetch a current page, or an installed tool may expose one provider's actions. The prompt and repository then add constraints such as the framework, traffic, region, budget, existing contracts and team skills.

The mix changes across products and sessions. A model-only answer may rely on older information, while a browsing tool can retrieve current documentation and an agent with account access can attempt a deployment. Outside observers generally cannot inspect the model weights or every retrieval and ranking decision, so a single answer cannot establish market share or a durable preference.

Google Search says its generative AI features use core Search ranking and quality systems to retrieve current pages from its index, and that eligible pages must be accessible and crawlable. Google also says meeting those conditions does not guarantee indexing or inclusion. That guidance explains Google Search, not every general-purpose model or coding agent.

The same Google guide says Search does not use llms.txt and that maintaining the file neither helps nor harms a site's Search visibility or rankings. Cloudflare's indexes can still help tools that choose to consume them. The distinction matters because a machine-readable entry point for one agent is not a universal recommendation signal.

Services need discoverable, executable and verifiable tasks

A service trying to enter AI-assisted decisions needs a narrower objective than brand exposure. It should define a task that a user can request, an agent can attempt and a system can judge. An accounting service might test the import of one sample invoice file, while an API provider might test one request with a documented response.

Publish the decision facts. State who the service is for, prerequisites, supported regions, inputs, outputs, current pricing boundaries, exclusions and the date of the documentation.
Document one complete path. Provide the minimum setup, sample data, expected result, cleanup steps and known failure modes for a task that can run in a sandbox or test account.
Expose stable machine interfaces. An OpenAPI description can define API operations, while a noninteractive CLI should return exit codes and structured output. An MCP server or agent skill should expose only capabilities whose underlying interface has already been tested.
Separate authority by risk. Public reads can require less control than creating records, while payment, deletion, publication and access to personal data need explicit permissions and an appropriate human approval point.
Make failures repairable. Error responses should identify the failed field or condition, explain the allowed correction and link to the relevant instruction. A stable error code gives an agent a safer next action than repeated guesses.
Supply evidence beyond marketing claims. Named customer cases, independent evaluations, security documentation, status history, refund terms and correction records help an agent verify that a stated capability exists under defined conditions.

Public documentation alone cannot complete a task, and an API alone does not tell an agent when the service is appropriate. The useful path connects eligibility facts, executable instructions, bounded permissions, machine-checkable results and current evidence. Each stage also needs an owner who can correct stale information or stop an unsafe operation.

Visibility and task completion require separate tests

A discovery test asks whether an AI system names a qualified service and describes its conditions correctly. An execution test asks whether the system can use official instructions and authorized tools to finish a bounded task. A service can score well on one and fail the other.

Fix the cases before testing. Write discovery prompts with explicit needs and task prompts with a known successful result, then keep the starting context consistent across candidates.
Record the environment. Log the date, model and version when available, retrieval setting, installed tools, account permissions and documentation snapshot for every run.
Score discovery. Record whether the service appeared, where it appeared, which reason was given, whether eligibility and limits were correct and whether cited evidence supported the answer.
Score execution. Measure completion, time to first success, manual interventions, permission requests, unsupported actions, cleanup and recovery after a controlled error.
Trace each failure. Separate missing retrieval, stale facts, incorrect qualification, tool failure, unclear errors and a required approval. A change should target the stage that failed.

Repeated runs across fresh sessions can reveal variation, but the result only applies to the tested models, tools, prompts and dates. A platform mention is evidence of visibility in that test, not proof of general preference. A completed deployment is evidence for that bounded workflow, not proof that the platform fits every production system.

Global regulators publish framework for medical AI software updates

APPI News Editorial — Mon, 24 Aug 2026 17:22:33 GMT

The International Medical Device Regulators Forum (IMDRF) published final principles for predetermined changes to medical-device software on August 6, 2026. The document gives regulators and manufacturers a common structure for reviewing software updates, including changes to artificial intelligence (AI) models, before those changes reach users.

A predetermined change control plan (PCCP) can let an authorized product change proceed without a separate regulatory submission in jurisdictions that accept the mechanism. It is not permission for a model to retrain and deploy anything it learns: changes must remain within the device's original intended use or purpose, meet prespecified acceptance criteria and follow the authorized plan.

A PCCP defines change before deployment

IMDRF defines a PCCP as a manufacturer-proposed plan that identifies specific software changes, the protocol for implementing and controlling them, and an assessment of their effects. The plan is an optional part of an initial regulatory submission or a later change submission, not a standing exemption from oversight.

The implementation method does not change that boundary. The 2026 document says a plan may specify automatic changes, manual changes requiring human action, or a combination of both, while each jurisdiction decides what its rules allow.

A model retrained with newer data can fit inside a PCCP when the change stays within the original intended use or purpose and the regulator has authorized the plan. A new clinical purpose or another change outside that boundary needs to follow the applicable route in each country.

Three records make a planned change reviewable

The IMDRF framework connects three records. Together they state what will change, how the change will be tested and deployed, and whether its expected benefits outweigh its risks.

Description of Changes: the planned modifications, their rationale, their boundaries and the device characteristics or performance that can be verified.
Change Plan: the data, evaluation methods, acceptance criteria, deployment steps and user communications for each modification.
Impact Assessment: the effects of each change and all changes together, including benefits, risks, mitigations and possible effects on cybersecurity or interoperability.

Traceability joins these records. Each proposed change should point to its verification and validation plan, while version control should identify which PCCP a regulator authorized and whether a later revision needs another authorization.

IMDRF says acceptance criteria should be quantitative, statistically sound, appropriate to the risk and clinically meaningful, and that a change should not be implemented when it fails those criteria. The evaluation plan can also identify relevant data practices, inputs and outputs, performance measures and statistical tests.

Drift can trigger a review but does not validate an update

Changes in incoming data or observed performance can trigger an investigation or retraining step. A drift signal does not establish that a replacement model performs safely in the intended population or works in the existing clinical workflow.

Current guidance from the US Food and Drug Administration (FDA) lists observed data drift as one possible retraining trigger and asks manufacturers to define data management, retraining, performance evaluation and update procedures. It also asks how training, tuning and test data will remain traceable across the original and modified versions.

The evidence has to separate the trigger from the acceptance decision. A review can record the source population, collection sites, equipment, reference standard, missing-data treatment and subgroup coverage, then test the modified model against criteria set before release.

The US guidance recommends comparing a changed device with both the original version and the most recently modified version. That comparison can show whether an update corrected the targeted problem while preserving performance requirements that were not meant to change.

Regulatory effect varies by country

The 2026 IMDRF document establishes a common reference, not a single international authorization. It says implementation varies across jurisdictions, does not replace national laws and regulations, and notes that some jurisdictions may not accept PCCPs for regulatory review.

The current US FDA final guidance was issued on August 18, 2025, after an original version dated December 4, 2024, and applies to AI-enabled devices in the US 510(k), De Novo and premarket approval pathways. When the FDA has authorized a PCCP, a manufacturer may implement a change covered by that plan without a separate US marketing submission, provided the authorized methods are followed.

A product's authorization and change route must therefore be checked market by market. A separate APPI News comparison maps how the United States, Taiwan and the European Union place AI medical devices into different regulatory systems.

Data exchange remains a separate acceptance test

Health Level Seven International defines Fast Healthcare Interoperability Resources (FHIR) as a standard for exchanging health information electronically, built from modular Resources. FHIR can carry patient, observation and diagnostic-report data between a hospital system and an AI device, but it does not authorize the software change or validate the model.

HL7 says FHIR is not a security protocol and leaves authentication and authorization to the surrounding system. A technically valid exchange can still carry the wrong code, unit, timestamp or patient context, so local mappings, permissions and the meaning of each input and output need separate tests after an interface or model version changes.

APPI News has separately examined how FHIR can improve hospital-data exchange while leaving data quality and access control to each implementation. The same distinction applies during a model update: interface conformance and clinical performance require different evidence.

Users need the current version and its limits

Joint 2024 principles from the US FDA, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency say device information should cover intended use, target populations, inputs, outputs, workflow effects, performance, limitations and change-management strategies. They also identify timely notification of software updates as a good practice.

The US FDA guidance makes that communication more specific for its jurisdiction. The agency recommends that labeling tell users when a device has an authorized PCCP, explain that updates may alter performance, inputs or use, and describe implemented changes through updated instructions or a version history.

Those documents place evidence and communication duties mainly on manufacturers and regulators. They do not create one worldwide rule assigning legal responsibility among a vendor, hospital and clinician when an updated model produces an error.

Hospitals still need a local release record

A regulatory authorization does not show that a new version has passed a hospital's interface and workflow checks. The broader APPI News medical AI deployment framework covers intended use, evidence, interoperability, human control and monitoring; a release record can apply those controls to one update.

Regulatory scope: the product authorization, PCCP status and permitted changes in each market where the version will be used.
Release identity: the model, software, dataset and interface versions, with the deployment date and affected sites.
Acceptance evidence: the test population, equipment, measures, prespecified thresholds, subgroup results and unresolved failures.
Workflow change: altered inputs, outputs, screens, training material, human review steps and fallback process.
Stop and recovery: the monitoring trigger, investigation owner, authority to suspend use and the prior version available for restoration.

This five-part release record is an APPI News synthesis of the cited documents, not an IMDRF certification scheme or a legal test. Its purpose is to distinguish a regulator's authorization of a planned change from a hospital's decision to activate that version in a particular system.

Published frameworks do not measure hospital adoption

APPI News could not find comparable official data on how many hospitals worldwide have deployed software changes under authorized PCCPs at the time of writing. The available documents describe regulatory structures, manufacturer evidence and recommended transparency, not adoption rates or clinical outcomes after updates.

A PCCP can shorten regulatory handling for a covered software change where a regulator accepts and authorizes it. The evidence chain still has to show what changed, which data and tests were used, how users were notified and what happens when the release misses its acceptance threshold.

Companies need workload tests before choosing an AI model

APPI News Editorial — Mon, 24 Aug 2026 17:22:33 GMT

Google's Gemini application programming interface (API) documentation, updated August 14, 2026, separates stable, preview, latest and experimental model versions. That distinction affects production planning, but it does not show which model will meet a company's quality threshold, response-time target or budget on its own data.

Artificial intelligence (AI) model selection starts by defining a measurable job and the cost of getting it wrong. A company can then run the same representative cases through each candidate. The comparison should cover the complete application rather than a model name and retain enough evidence to repeat the decision after a version or price change.

Model catalogs define a shortlist, not a winner

A model is one component of an application. The application also includes instructions, retrieval, tools, access controls, output validation, logging and human review. A strong general model can still fail a narrow workflow if the surrounding system supplies the wrong documents, accepts malformed output or gives the model authority that the task does not require.

Provider catalogs can narrow the field by modality and operating constraints. OpenAI's current catalog distinguishes flagship, balanced and cost-sensitive models and lists separate offerings for image, real-time, speech, transcription and embedding work. The page also publishes model-specific context, output, tool and price information, all of which may change after this report's publication date.

Anthropic's model-selection guide frames the choice around capability, speed and cost and tells developers to test candidates with use-case-specific benchmarks, actual prompts and actual data. The guide then calls for comparisons of accuracy, response quality, edge cases and performance-cost tradeoffs.

Google's Gemini documentation identifies stable versions as the usual choice for production and warns that preview, latest and experimental endpoints carry different change, notice and availability conditions. Those lifecycle labels affect reproducibility and the amount of regression testing a deployment will need.

Those documents support a shortlist, not a neutral ranking. Each provider defines its own descriptions, tests and service boundaries. Buyers need to confirm current specifications on the date of a decision and then test the candidates against the same acceptance criteria.

Define the job and its failure cost first

Replace a broad objective such as “use AI for customer support” with an output that can pass or fail. A support-summary task might require a fixed set of fields, correct product names, no invented refund promise and completion within a defined time. A document-classification task needs an agreed label set and rules for ambiguous or incomplete records.

The failure cost determines the order of the gates. A low-impact internal draft may tolerate occasional correction if it is inexpensive and fast. A system that influences payments, employment, access to services or safety needs stricter quality thresholds, traceable evidence, limited automation and a tested route to human review.

Write the acceptance rules before testing. Otherwise, a team can lower a quality threshold after seeing a favored model's results or give one candidate a better prompt and more retries. Pass-or-fail gates should cover the output quality, prohibited behavior, data handling, maximum latency and operational constraints that the application cannot trade away.

The hardware decision sits below the same workload definition. Teams comparing local or private deployments can use the site's workload-based AI accelerator checklist to measure memory, throughput, tail latency, power and full-system cost after they have identified model candidates.

Build a test set from production work

A useful test set reflects the inputs that the service will actually receive. Include common cases, long and short inputs, supported languages, tables, spelling errors, missing fields, conflicting instructions and rare cases with high failure costs. Keep a separate holdout set so repeated prompt changes do not overfit the evaluation cases.

Run each candidate with the same input, system instructions, retrieval collection, tools, sampling settings and scoring method. If a provider requires a different implementation, document the change and decide whether it represents a necessary production advantage or an unfair test condition. Keep the model identifier and test date because an alias can point to different behavior later.

Score task success before style. Depending on the job, that can mean exact field accuracy, classification precision and recall, citation support, execution success or a blinded reviewer rubric. Record refusals and unsupported claims separately from formatting errors because they require different controls.

Generative models can return different answers to the same input. Repeat a sample enough times to reveal variation, especially for high-cost cases, and preserve every failed output. The internal article on why AI systems produce fluent unsupported answers explains why a confident response does not remove the need to check the underlying source.

Measure latency and cost at the application level

Token prices do not equal the cost of a successful business result. A cheaper model may require more retries, a longer prompt, extra retrieval calls or more human correction. A higher-priced model may still cost less per accepted document if it completes the job on the first attempt.

For every candidate, record input and output volume, tool calls, retries, cache behavior, successful outputs and reviewer time. Divide the full run cost by accepted outputs, not total requests. Use normal and peak traffic because rate limits, queuing and concurrency can change the result.

Average response time can conceal a poor user experience. Interactive tests should include median latency and a high percentile such as the 95th or 99th, while streaming applications may also need time to first output. Batch work needs completion time and throughput under the intended volume rather than an interactive-response measure.

Forecast a range instead of one annual total. Input length, output length, traffic, exchange rates and provider prices can all change. Local deployment adds hardware, electricity, cooling, monitoring, upgrades and staff time; a cloud application adds network transfer, observability and any paid retrieval or tool service.

Data controls can eliminate a model before quality scoring

Inventory the information the application will send, retrieve, generate and log. Mark personal data, confidential business material, licensed content and records subject to contractual or legal restrictions. That inventory should include prompts, uploaded files, embeddings, retrieved passages, tool results, feedback and support logs.

Then verify the exact service and contract, not the provider's brand in general. Consumer chat products, self-service APIs and negotiated enterprise services can use different retention, training, administrator and regional-processing settings. Public model pages do not establish which controls apply to a particular account.

A deployment claim also needs technical evidence. Test access permissions, deletion, logging and incident procedures in the proposed environment. If data must remain in a defined region or network, obtain contractual language and architecture records that identify every processor and transfer path.

Versioning and fallback plans belong in the evaluation

Version labels can carry operational consequences. Google's stable identifier points to a specific model that usually does not change, while its latest alias can be replaced by a newer release. Its preview and experimental categories can have tighter limits or less stable availability.

Record the exact identifier, provider region, parameters, system instructions, tool definitions, retrieval index and evaluation set for every approved release. Run regression tests before changing any of them. A provider announcement that a new model scores better on a general benchmark does not show that it preserves a company's required format or handles its edge cases.

Keep a second model or a non-AI procedure for service interruption, price changes, capacity limits and quality regression. The fallback needs its own acceptance result and operating instructions. An untested model name in a continuity plan is not a working alternative.

Risk frameworks turn model choice into lifecycle management

The US National Institute of Standards and Technology says its AI Risk Management Framework is voluntary and applies to the design, development, use and evaluation of AI products, services and systems. NIST released version 1.0 on January 26, 2023, and says a revision is now in progress.

The framework's Core organizes work into Govern, Map, Measure and Manage and calls for continuous risk management throughout the AI system lifecycle. NIST says these functions are not a checklist or a required sequence. Organizations apply them according to context, resources and risk.

The International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) published ISO/IEC 42001:2023 in December 2023 to specify requirements for establishing, implementing, maintaining and continually improving an artificial intelligence management system. It covers organizations that provide or use AI, but it does not rank individual models or establish that one is appropriate for a specific task.

For model selection, Govern assigns ownership, approval authority, documentation and third-party controls. Map defines the intended users, operating context, benefits, failure costs and human oversight. Measure covers the representative tests, uncertainty, independent review and monitoring, while Manage decides whether deployment should proceed and how to respond, disengage or deactivate when performance falls outside the approved range.

These frameworks keep model quality from becoming the only decision. A candidate may produce the best sample answers and still fail because the organization cannot meet its data obligations, recover from an outage or monitor changes. The selected model needs both an acceptance record and an operating control plan. Neither framework replaces the laws and contracts that apply in each jurisdiction where a system is developed or used.

A six-part decision record makes the result repeatable

Task and boundaries: Define the allowed inputs, required output, users, prohibited decisions and human escalation path.
Data inventory: Record each data class, permitted location, retention rule, access role and third party that may process it.
Acceptance tests: Use representative cases and fixed scoring for quality, unsupported output, latency, unit cost and human correction.
Service terms: Confirm prices, limits, version policy, training use, deletion, regional processing, support and deprecation notice for the contracted service.
Operations: Assign monitoring, incident response, review cadence, model-change approval and the authority to stop the system.
Exit plan: Test a fallback model or manual route and estimate the work needed to move prompts, evaluations, data and integrations.

The decision should identify which candidates failed mandatory gates and show the raw results for those that passed. A weighted score can compare cost, speed and operational effort after the gates, but it should not allow a low price to offset a privacy or safety failure.

Set review triggers rather than relying only on a calendar. A new model version, price change, altered data distribution, sustained traffic increase, new error pattern or change in legal duties should start another controlled run. Reusing the same decision record makes that reassessment faster and shows why the approved choice changed.

Frequently asked questions

Is the largest model usually the best choice?
No. A smaller model can win a constrained, high-volume task if it meets the quality gate with lower latency and cost. Complex reasoning or multimodal work may justify a higher-capability model, but the same production test must establish that advantage.

Should a company choose an API, private cloud or local model?
Start with data-location requirements, latency, traffic, available hardware and operating skills. An API can reduce infrastructure work, while a private or local deployment can offer more direct control over the environment but shifts updates, capacity, monitoring and failures to the buyer.

Can a public leaderboard select the model?
No. A benchmark can indicate a capability under its own dataset and scoring method. It does not measure a company's prompts, languages, retrieval system, failure costs, contract or production traffic.

How often should a model be reassessed?
Reassess when a version, price, workload, data distribution, failure pattern or legal requirement changes. Higher-impact systems also need scheduled reviews and continuing monitoring, with a tested route to human handling or shutdown.

Growth and swallowing signs guide care for picky-eating toddlers

APPI News Editorial — Mon, 24 Aug 2026 16:05:56 GMT

The World Health Organization (WHO) updated its infant and young child feeding fact sheet on August 4, 2026, and recommends responsive feeding that follows hunger and fullness cues for children aged six to 23 months. For a toddler refusing vegetables, the practical question is whether the refusal is a short-lived preference or occurs with poor growth, a very narrow diet, or signs of feeding and swallowing trouble.

The US-based American Academy of Pediatrics (AAP) says many children begin to resist unfamiliar foods between 18 months and age two. Its guidance pairs a familiar food with small portions of other foods and lets the child decide what and how much to eat from what is offered.

Breathing and swallowing signs come before meal strategy

The US-based American Speech-Language-Hearing Association (ASHA) lists repeated gagging or coughing, slow or ineffective chewing, food left in the mouth, a wet-sounding voice, congestion during or after eating, and restricted dietary variety among possible feeding or swallowing disorder signs. Vomiting, constipation, poor weight gain and failure to move toward age-appropriate textures add reasons for a clinical assessment.

Children’s Health Queensland in Australia says a child who cannot swallow because food appears stuck needs emergency-department care, while a blockage that impairs breathing requires an ambulance. Caregivers should call the emergency service where the child is located rather than waiting to see whether the next meal goes better.

The same Australian guidance calls for medical review when swallowing difficulty recurs or appears with weight loss, regurgitation or vomiting. Repeated coughing or choking, blue color around the lips, rapid breathing, a wet or gurgly voice, or abnormal pauses in breathing during meals also point beyond ordinary food preference.

England’s National Health Service (NHS) lists drowsiness, reduced urination or fewer wet diapers, few or no tears, and fast breathing among signs of serious dehydration that need urgent treatment; difficulty breathing or being difficult to wake requires emergency care. A toddler who refuses food but drinks normally and remains active differs from a child who is no longer taking fluids or passing urine.

England’s NHS says sudden breathing problems or swelling of the throat or tongue after a food can signal anaphylaxis, which requires emergency treatment. That reaction should not be handled as ordinary food refusal or tested by serving the food again at home.

Responsive feeding replaces pressure at the table

WHO’s 2023 global guideline recommends responsive feeding for children aged six to 23 months, defining it as support for autonomous eating in response to physiological and developmental needs. WHO labels the recommendation strong but rates the certainty of evidence as low, so it should not be presented as a guaranteed way to end picky eating.

The AAP advises regular meal and snack times, a shared family meal when possible, and at least one food the child already accepts. Adults decide what food is offered; the child decides whether and how much to eat, without a separate replacement meal after a refusal.

Rejected food can return at a later meal in a small portion, and repeated offers may be needed before a child accepts it. Pressure, punishment and dessert as payment for eating vegetables can intensify conflict, while adults eating the same foods provide a visible model.

A pattern across days says more than one refused food

England’s NHS advises looking at what a toddler eats across a week rather than judging one day; being active, gaining weight and appearing well are reassuring signs. Those signs do not rule out a feeding problem when swallowing symptoms or severely restricted variety persist.

WHO’s August 2026 fact sheet advises three to four meals a day for children aged nine to 23 months, with one to two additional nutritious snacks as required. It also calls for food consistency and variety to increase gradually while caregivers respond to hunger and fullness cues.

One untouched vegetable does not show whether a toddler’s overall diet supplies enough energy and nutrients. The more useful pattern includes the range of foods and textures accepted, whether whole food groups have dropped out, how the child manages chewing and swallowing, and whether growth continues as expected.

Growth and a short meal record guide the next assessment

WHO’s growth assessment guidance says health workers should measure a child’s weight and length or height, compare the measurements with growth standards, and plot them to identify a growth problem or trend. A single percentile or one meal cannot establish why a child is eating less.

Before an appointment, caregivers can keep a record of meals, snacks and drinks, including what was offered and what the child accepted. Notes on meal duration, textures, coughing, gagging, food remaining in the mouth, vomiting, constipation and recent growth measurements help a clinician see whether the main concern is dietary range, feeding skill, swallowing, gastrointestinal symptoms or growth.

ASHA says a feeding and swallowing evaluation can review dietary diversity, intake, chewing, swallowing, mealtime routines, the eating environment and caregiver-child interaction. Depending on the findings, a pediatrician may involve a dietitian, speech-language pathologist or another member of a feeding team.

Children born preterm and those with chronic heart, lung, neurological or gastrointestinal conditions, developmental differences, known food allergy, or a previous feeding or swallowing problem need advice tailored to their history. General strategies for an otherwise well toddler do not establish which foods or textures are safe for these groups.

England’s NHS advises families of a child with a diagnosed food allergy or eczema to seek clinical guidance when introducing foods and warns against removing a major food such as milk without professional advice because the child may miss needed nutrients. A diagnosed allergy plan takes precedence over general repeat-exposure advice.

APPI News could not find a single internationally accepted numeric threshold for how few foods defines a toddler feeding disorder at the time of writing. The official sources instead assess restricted variety alongside nutritional adequacy, feeding skills, symptoms and growth.

Common questions

Does one day without vegetables require medical care?
Not on that fact alone. Review the child’s intake across several days, activity, general condition and growth, while acting sooner if breathing, swallowing or other warning signs appear.

Should a toddler be made to finish the vegetables on the plate?
Guidance from WHO, the US-based AAP and England’s NHS favors calm, repeated offers and attention to hunger and fullness cues over pressure or food rewards. A familiar item can remain on the plate beside a small amount of the less familiar food.

Which mealtime signs need an assessment?
Repeated coughing, choking or gagging, slow chewing, food remaining in the mouth, a wet-sounding voice, breathing changes, persistent vomiting or constipation, a very narrow diet, or a concerning growth trend all merit discussion with a clinician. Inability to swallow or breathing obstruction is an emergency rather than a routine picky-eating concern.

Japanese trial finds fewer advanced cancers with added ultrasound

APPI News Editorial — Mon, 24 Aug 2026 12:00:00 GMT

A Japanese randomized trial published in The Lancet in February 2026 enrolled 72,661 asymptomatic women ages 40 to 49 and found a lower cumulative incidence of stage II or higher breast cancer after two rounds of mammography plus ultrasound than after mammography alone. Median follow-up was about 11 years.

The result adds long-term evidence for supplemental ultrasound, but it does not create a worldwide screening rule. The study did not use breast cancer mortality as an endpoint, and its age range, Japanese setting and two-round design limit how far the result can be applied.

New breast changes need prompt assessment

Screening is intended for people without symptoms. The World Health Organization (WHO) lists a new breast lump or thickening, a change in breast size or shape, skin dimpling or redness, a nipple or areola change, and abnormal or bloody nipple fluid as possible breast cancer symptoms.

A person with one of these changes should seek clinical assessment promptly rather than wait for a screening invitation or the next routine mammogram. Most breast lumps are not cancer, so a symptom is not a diagnosis; it is a reason to move from routine screening to diagnostic assessment.

Mammography evidence does not produce one global schedule

The International Agency for Research on Cancer (IARC) found sufficient evidence that mammography screening reduces breast cancer mortality among women ages 50 to 69. Its 2016 evaluation estimated an average 24 percent reduction among women invited to service screening and about 40 percent among those who attended. The gap shows how the population counted changes the result.

WHO describes screening as mammography in an apparently healthy population, usually women ages 50 to 69. The US Preventive Services Task Force instead recommends mammography every two years from ages 40 to 74 for people at average risk in the United States.

Neither range is a worldwide eligibility or coverage rule. The applicable starting age, interval and route to follow-up come from the program or clinical guidance used in the country where care is received.

Mammography also has harms. The US task force lists false-positive and false-negative results, extra imaging and biopsy, overdiagnosis and radiation exposure; IARC found that the radiation-related cancer risk was outweighed by the mortality benefit for women ages 50 to 74, but that conclusion does not erase the other costs of screening.

Density describes the mammogram, not a disease

The US National Cancer Institute (NCI) defines breast density as the relative amount of fibrous, glandular and fatty tissue seen on a mammogram and says density cannot be determined by touch. In reports using the Breast Imaging Reporting and Data System (BI-RADS), the two highest categories, C and D, are described as dense.

Fibrous and glandular tissue appears white on a mammogram, as do some tumors and other abnormalities, while fatty tissue appears dark. This overlap makes mammography less sensitive in dense breasts, and NCI also identifies density as a breast cancer risk factor separate from the masking effect.

Dense breasts are not an abnormal condition or a cancer diagnosis. Density can change with age and other factors, and a radiologist's classification is an assessment at the time of the mammogram.

A dense result does not make mammography pointless, and a negative mammogram does not guarantee that no cancer is present. The decision on added imaging depends on the complete report, the person's wider risk profile and the guidance in the health system providing care.

Ultrasound finds more cancers and more false positives

The 2026 J-START analysis reported 234 stage II or higher cancers in the ultrasound group and 277 in the mammography-only group. The hazard ratio was 0.83, with a 95.6 percent confidence interval of 0.70 to 0.98.

The difference between the trial groups was statistically distinguishable only between months 48 and 96, and the curves then remained stable. The trial enrolled women without a breast-density entry requirement, covered only ages 40 to 49 in Japan and did not show whether the lower incidence of advanced cancer will translate into fewer deaths.

A 2023 Cochrane review found that adding ultrasound detected two more breast cancers per 1,000 women over two years, but produced 37 more false-positive results among every 1,000 women without cancer and led to 27 more biopsies per 1,000 women screened. The review searched through May 2021 and found no study that established a reduction in breast cancer or all-cause mortality.

In 2024, the US task force found insufficient evidence to recommend for or against supplemental ultrasound or magnetic resonance imaging (MRI) after a negative mammogram in people with dense breasts. The European Commission Initiative on Breast Cancer conditionally advised organized screening programs against routine automated or hand-held supplemental ultrasound for asymptomatic women with dense breasts and a negative mammogram, citing low or very low certainty. These positions do not mean that supplemental imaging is never used; they mean that density alone has not produced a universal routine schedule.

Higher-risk and special populations need separate pathways

The average-risk schedules above do not cover every person. The US task force excludes people with a high-risk genetic marker or syndrome, high-dose chest radiation at a young age, previous breast cancer or a high-risk breast lesion from its average-risk recommendation.

The adult screening ranges cited here do not apply to children and do not establish one schedule during pregnancy or breastfeeding. People in those groups should use clinician-led assessment if a breast symptom appears rather than apply an adult routine screening interval.

NCI says postmenopausal hormone therapy is associated with higher breast density. The cited sources do not provide a general basis for changing a prescribed medicine solely because a mammogram reports dense tissue; medicine changes and screening decisions require clinical judgment.

A dense result changes the questions, not the diagnosis

A review of a mammogram report should separate the density category from the radiologist's assessment and recommended follow-up. Density alone is neither a cancer diagnosis nor an automatic order for ultrasound.

After a negative mammogram, the case for another imaging test depends on total breast cancer risk, the density category, prior findings and local guidance. An abnormal screening result or a new symptom follows a diagnostic pathway instead of the next routine screening interval.

The source article presented a 41 percent mortality reduction as a general screening effect. IARC's closest international evaluation produced different estimates for invitation and attendance within specified age groups, so APPI News has not treated 41 percent as a worldwide personal benefit. The source's stage-specific survival percentages were also omitted because no international dataset with matching stage definitions and population was identified.

Frequently asked questions

Does a dense-breast result mean cancer?
No. It describes the mix of tissue seen on a mammogram, although it can both mask cancers and raise breast cancer risk.

Can breast density be identified by touch?
No. NCI says a radiologist determines density from a mammogram, not from a self-exam or clinical breast exam.

Does dense tissue make mammography useless?
No. Mammography remains the core population screening test, but it is more likely to miss a cancer when tissue is dense.

Should everyone with dense breasts add ultrasound?
No universal rule supports routine ultrasound for everyone with a dense result and an otherwise negative mammogram. The potential for added detection has to be weighed against false positives, more imaging and biopsy.

What screening age and interval apply?
There is no single worldwide schedule. Eligibility and intervals should come from the program or clinical guidance used where the person receives care.

Should a new lump wait until the next mammogram?
No. A new lump, nipple change, abnormal nipple fluid or skin change needs clinical assessment rather than waiting for routine screening.

Microsoft documents Windows driver cleanup but no universal 30-day rule

APPI News Editorial — Mon, 24 Aug 2026 11:06:04 GMT

Microsoft's Windows troubleshooting page, updated February 12, 2026, documents a past Windows 10 incident in which an update reliability tool removed third-party USB drivers from the INF cache. AWS separately ties a 30-day Plug and Play Cleanup scan to Windows Server 2012 R2 images released before September 10, 2014, not to all current Windows PCs.

The documents establish that cleanup mechanisms have caused failures in narrow conditions. They do not support a universal rule that Windows 10 or Windows 11 deletes any device driver after 30 idle days, and they do not establish that ordinary inactivity caused a missing graphics driver.

What the documented cleanup cases establish

Microsoft's case involved a Windows 10 Long-Term Servicing Branch application that depended on custom or third-party USB drivers remaining in the INF cache. Process Monitor showed the update reliability tool launching pnpclean.dll with /DEVICES /DRIVERS /MAXCLEAN, after which the application could no longer find those drivers. Microsoft says the tool's applicability rules were improved and version 10.0.14393.10020 or later should not cause the documented issue.

AWS says Plug and Play Cleanup scanned for inactive system devices every 30 days and mistakenly removed the EC2 network device on Windows Server 2012 R2 Amazon Machine Images released before September 10, 2014. AWS says the problem occurred only with images made available before September 2014 and recommends a driver upgrade or its remediation script for those affected instances. This is evidence about an old server image, not a timing rule for present-day consumer laptops and desktops.

Idle maintenance is not the same as driver cleanup

Microsoft's Automatic Maintenance compatibility page says maintenance tasks can run at scheduled times or while a PC is idle and can pause when the user resumes activity. The page lists Windows 8 and Windows Server 2012 as its platforms. It does not identify a 30-day driver-deletion policy for Windows 10 or Windows 11.

Timing alone cannot show why a device entry changed. Record the Windows edition and build, recent updates, exact hardware model and any Device Manager status code before uninstalling anything. APPI News's Windows 11 memory guide separates process, cache, kernel and driver allocations before software changes are made, because a high memory percentage is a separate symptom rather than evidence that a driver was removed.

Check the device and package separately

Open Device Manager and look under the relevant hardware category. Microsoft's Windows 10 and 11 procedure directs users to select View > Show hidden devices before uninstalling a device or package. If an entry is present, record its name, status and driver details; a hidden entry alone does not prove cleanup deleted its package.

Microsoft's driver inventory guidance uses PnPUtil to enumerate third-party packages and identify packages with no devices attached. An elevated Command Prompt can run pnputil /enum-drivers for an initial list. Inventory is diagnostic; it is not a reason to delete every package that appears unused.

Microsoft distinguishes uninstalling a device from deleting its driver package: uninstalling removes the device-driver association but leaves the package in the Driver Store, where Plug and Play can use it again if the device is detected. Deleting the package removes it as an installation source, and Microsoft says a later reinstall then requires the original source, such as Windows Update. Manually deleting files inside the Driver Store can produce unpredictable behavior.

Reinstall before considering package deletion

Identify the exact device before choosing a package. A listing of Microsoft Basic Display Adapter means Windows is using its built-in generic driver instead of a manufacturer-specific driver. Microsoft recommends checking Windows Update first and then the display-adapter manufacturer's official site if Windows Update has no suitable package.

Save work and use the supported reinstall path. Microsoft tells Windows 10 and 11 users to uninstall the selected device in Device Manager, restart, and let Windows attempt to reinstall the driver. If that attempt fails, the same page directs users to Windows Update or a compatible driver downloaded from the hardware manufacturer's official site.

Match the device, operating system and architecture. Microsoft says a manually downloaded package must match the Windows version and architecture, and it warns against driver sites that are not operated by the manufacturer. Organization-managed computers may also have controlled packages and policies, so their administrators should handle the change.

Stop after service is restored. A working manufacturer driver and normal Device Manager status remove the need for package deletion. If the device still fails, record the status code and use the computer or hardware manufacturer's support channel rather than deleting unknown INF files.

Do not copy cleanup or registry commands

The Microsoft USB incident records pnpclean.dll /DEVICES /DRIVERS /MAXCLEAN as a command launched by an update tool; it does not present that command as a general maintenance procedure for users. Running it manually would invoke the documented cleanup routine without establishing which packages are safe to remove.

APPI News could not verify in current Microsoft documentation that changing the Device Driver Packages Autorun registry value to 0 is a supported general remedy for current Windows 10 or Windows 11 systems. The registry change is therefore omitted from this guide. Microsoft instead says to verify that a package is not needed, test removal on a noncritical PC, keep backups and establish a restore point before deleting a package.

Frequently asked questions

Does Windows delete every unused driver after 30 days?
No. AWS documents that interval only for old Windows Server 2012 R2 EC2 images, and the cited Microsoft sources do not establish it as a general Windows 10 or Windows 11 policy.

Does Microsoft Basic Display Adapter prove that cleanup ran?
No. It shows that Windows is using a generic driver because a manufacturer-specific driver is not available or installed; it does not identify why that driver is missing.

Can PnPUtil delete a suspected package?
Yes, but Microsoft says the package must first be identified and checked against the devices that use it. Removing an essential package can cause instability or data loss, so deletion is not a diagnostic first step.

Should the Device Driver Packages Autorun value be set to 0?
APPI News could not verify that change as a current, general Microsoft recommendation for Windows 10 or Windows 11. This guide does not recommend the registry edit.

US data center opposition rises as Taiwan reviews power rules

APPI News Editorial — Mon, 24 Aug 2026 09:06:18 GMT

The University of Pennsylvania's Annenberg Public Policy Center reported on August 11, 2026, that 61 percent of US adults somewhat or strongly opposed construction of new data centers in their area, up from 49 percent in a spring survey. The poll measured opposition, but it did not ask respondents to select water use, electricity demand, noise or any other reason for their answer.

Four months of polling therefore show a change in US opinion, not the cause of that change. The June 16 to July 19 survey covered 1,320 US adult citizens, was weighted to national population benchmarks and had a margin of error of plus or minus 3.5 percentage points.

Texas turns resource questions into project disclosures

Governor Greg Abbott of the US state of Texas directed the Public Utility Commission of Texas and the Electric Reliability Council of Texas (ERCOT) on August 3 to audit every data center advancing through ERCOT's grid-interconnection process before those projects move forward. The governor's office said a project that failed to meet the two bodies' requirements would be denied a connection to the Texas grid.

The directive asks each project for annual and peak electricity demand, on-site generation, annual and peak water demand, water sources, reuse plans and cooling technology. It also covers tax incentives and other public support, ownership, noise and light controls, traffic, setbacks and emergency coordination.

This is narrower than a statewide construction ban. The directive addresses projects in ERCOT's grid-interconnection queue rather than every land-use or building permit issued by a local government.

National estimates cannot predict one site's impact

Lawrence Berkeley National Laboratory's 2024 model puts direct water consumption by all US data centers in 2028 at about 145 billion to 275 billion liters, or 145 million to 275 million cubic meters. The range comes from scenarios for equipment shipments, operating patterns, facility types, cooling systems and their efficiency; it is not a sum of meter readings from future facilities.

The laboratory's update published in June 2026 estimates that data centers could account for 11.8 percent of total US electricity use in 2030, with scenarios ranging from 9.5 percent to 15.3 percent. Its reference case is 649 terawatt-hours, while compounded uncertainty scenarios span 521 to 843 terawatt-hours.

Both reports describe national scenarios. A proposed site's demand still depends on its installed hardware, utilization, operating hours, climate, cooling configuration and power supply, so a national average cannot establish whether a particular community has enough grid or water capacity.

Cooling choices move demand between water and power

Guidance from the US Department of Energy's Federal Energy Management Program says cooling-tower consumption comes mainly from evaporation and blowdown, the water drained to control dissolved minerals. The guidance says higher cycles of concentration can reduce blowdown, while air-side and water-side economizers can cut water and energy use when climate and system design allow them to operate.

Other measures carry their own loads. The same guidance says reverse osmosis can return treated blowdown to a cooling tower but increases energy demand and maintenance, while direct liquid cooling still needs a final path for rejecting heat and may remain connected to a cooling tower.

A project record should therefore identify the complete heat-rejection system, not stop at labels such as air cooled, liquid cooled or closed loop. Annual consumption, peak-day demand, source water, reuse, discharge and the response to dry conditions belong beside normal and peak electricity demand.

Taiwan has a current energy rule and a separate draft

Taiwan's Energy Administration is the central authority responsible for the country's energy policy and energy supply-and-demand planning. The agency says hyperscale and colocation data centers with at least 5 MW of contracted electricity capacity must submit an Energy Utilization Manual before construction or expansion and adopt best available techniques. Its English-language summary sets power usage effectiveness ceilings of 1.3 for hyperscale facilities and 1.4 for colocation facilities.

The ministry's 2025 national electricity report includes added demand from data centers and AI technology in its forecast of about 2.5 percent average annual electricity-demand growth from 2026 through 2035. That forecast also incorporates economic growth, temperature, semiconductor expansion and expected energy savings, so 2.5 percent is not a data-center-only growth rate.

On July 1, 2026, Taiwan's Ministry of Economic Affairs proposed a separate amendment that would add five industry-benefit categories to the energy review: investment and output, employment, supply-chain effects, support for the AI sector, and information security and resilience. The proposal would also bring the host local authority or industrial-park authority into the review before the file reaches the Energy Administration.

Taiwan's official English law database still identified November 3, 2025, as the latest revision of the regulations when checked on August 24, 2026. APPI News did not find a later promulgation or an official English version of the July proposal, so the five industry-benefit categories remain a draft rather than a current legal requirement.

A review record should separate forecasts from commitments

The US and Taiwan records use different legal processes, but they expose the same information problem: a nameplate capacity or national forecast does not show how one project will affect one grid connection, water system or public budget. A project-level file can make those differences visible without assuming that every data center has the same footprint.

Electricity: Record contracted and on-site capacity, expected annual use, normal and peak demand, the grid-connection point, the connection date and which upgrades are required.
Cooling and water: Identify the cooling and final heat-rejection systems, annual and peak water demand, water sources, reuse, discharge and operating changes during dry periods.
Public costs: List tax incentives, grants and other public support, then state who pays for grid, water, road and emergency-service upgrades.
Community effects: Provide measurable plans for noise, light, setbacks, traffic and emergency response, with the party responsible for each commitment.
Local benefits: Distinguish projected investment, employment and supply-chain activity from binding commitments, and state how each result will be measured.

Facility demand also has to be built from the proposed equipment rather than an inherited server-room average. APPI News's AI rack power planning guide explains how peak rack load, redundant feeds and cooling equipment change the site-level capacity record.

Frequently asked questions

Does the 61 percent survey result show why US residents oppose data centers?
No. The Annenberg survey measured support and opposition to new data centers in respondents' areas, but it did not ask them to rank water, electricity, noise, traffic or other possible reasons.

Does liquid cooling eliminate data center water use?
No single answer applies to every design. A liquid loop can move heat efficiently inside a facility, but site water demand depends on the final heat-rejection equipment, climate, operating conditions and whether water is reused.

Is Taiwan's five-part industry-benefit test already in force?
APPI News could not verify that it had been promulgated by August 24, 2026. Taiwan's existing energy-review and efficiency requirements for qualifying data centers remain separate from the July proposal.

Google launches four Pixel 11 models in Taiwan from about US$941

APPI News Editorial — Mon, 24 Aug 2026 00:00:00 GMT

Google began selling four Pixel 11 models in Taiwan on August 20, 2026, with the standard phone starting at NT$29,990 (about US$941). The Pixel 11 Pro and Pro XL add brighter displays and a 120x zoom setting, while their 256GB configurations have 12GB of random-access memory (RAM) and Google has not documented Suica support for Taiwan-sold units.

The launch covers a standard phone, two conventional Pro models and a foldable. Google’s local announcements and global specifications establish their prices and hardware, but they do not support treating an unofficial Japanese transit-card setup as a promised Taiwan feature.

Four Taiwan prices span US$941 to US$1,915

Google’s Taiwan launch notice says the Pixel 11, Pixel 11 Pro and Pixel 11 Pro XL went on sale on August 20 at NT$29,990, NT$36,990 and NT$42,990, respectively, with storage starting at 256GB. Using Taiwan’s August 24 closing rate of NT$31.854 per US$1, those prices equal about US$941, US$1,161 and US$1,350.

Google’s separate Taiwan announcement prices the Pixel 11 Pro Fold from NT$60,990 and lists 256GB, 512GB and a Google Store-exclusive 1TB option. At the cited exchange rate, that starting price is about US$1,915, slightly more than twice the price of the standard Pixel 11.

These are Taiwan list prices, not global price conversions or forecasts for another market. Retail promotions, mobile-service contracts, warranty terms and available storage can differ by country and sales channel.

The Pro premium buys display and camera hardware

Google lists the standard Pixel 11 with a 160 mm Actua display rated at up to 3,000 nits, 12GB of RAM, a 48-megapixel main camera and a 5x optical telephoto camera with Super Zoom up to 30x. It uses the same Tensor G6 processor and Titan M3 security coprocessor named across the four models.

The Pixel 11 Pro and Pro XL specifications raise peak brightness to 3,600 nits and list a 50-megapixel main camera, 48-megapixel ultrawide camera and 48-megapixel 5x telephoto camera. Google specifies optical-quality output at 10x and Pro Zoom up to 120x, so the 120x headline is not a claim of 120x optical magnification.

The smaller Pro has a 161 mm display and weighs 204 g, while the Pro XL has a 171 mm display and weighs 226 g. Their rear-camera specifications match, while screen size, weight, battery capacity and charging are among their documented differences.

The Pixel 11 Pro Fold has a 164 mm cover display, a 204 mm folding display, 16GB of RAM and Super Zoom up to 30x. Its higher price buys the folding form factor and larger inner screen, not the Pro and Pro XL’s 120x zoom setting.

The 256GB Pro models use 12GB of RAM

Google’s specifications assign 12GB of RAM to the 256GB Pixel 11 Pro and Pro XL, then 16GB to their 512GB and 1TB versions. The standard Pixel 11 has 12GB at both listed storage capacities, while the Pro Fold has 16GB across its storage range.

The same Google page lists 16GB of RAM for every Pixel 10 Pro and Pro XL storage option. On Google’s published specifications, the 256GB Pixel 11 Pro and Pro XL therefore have 4GB less RAM than the directly preceding Pro models, while storage now starts at 256GB.

Storage and RAM are not interchangeable: storage holds apps and files, while RAM affects active workloads and background retention. APPI News’s phone RAM guide recommends testing app reloads and checking documented on-device artificial intelligence (AI) requirements instead of treating 16GB as a universal speed threshold.

Google does not document Suica on Taiwan-sold units

The Suica question concerns a Japanese transit and stored-value card, not a payment service native to Taiwan. A phone’s contactless radio does not by itself establish that the device can enroll a particular transport card.

Google Wallet’s Taiwan support table lists iPASS, a Taiwanese stored-value payment service, and supported bank cards; it sends readers to Suica and PASMO instructions explicitly labeled “Japan only”. The table does not list EasyCard, Suica, PASMO or ICOCA as supported Taiwan e-money.

Google’s 2026 Pixel 11 specifications list near-field communication (NFC), but APPI News could not find a Taiwan-specific FeliCa or Osaifu-Keitai certification on that page. NFC support alone is not evidence that Google Wallet will offer Suica enrollment on a Taiwan retail unit.

APPI News also could not find official Google or East Japan Railway Company instructions that validate changing the Google Play region, installing a Japanese SIM or using a particular foreign credit card as a supported Taiwan-unit setup. Those steps are omitted because an unofficial workaround can stop working and does not create a product commitment.

Gemini features are not uniform across markets

Google says Tensor G6 runs Gemini Nano and powers new Pixel 11 assistance and camera features, but its announcement limits availability by country and language, requires users to be at least 18 for some features and says higher usage may require a subscription. The processor name therefore does not prove that every demonstrated function is available in Taiwan or any other market.

Google also bases its advertised browsing, app-launch and AI-efficiency gains on internal comparisons with Tensor G5, including tests on preproduction devices. Those figures describe Google’s test conditions, not an independent retail benchmark or a guaranteed improvement in every app.

Seven years of updates start with US availability

Google’s support policy names all four Pixel 11 models and promises seven years of operating-system and security updates from the date each phone first became available on the Google Store in the United States. The period may include new or upgraded Pixel Drop features, but the policy does not promise that every feature will reach every market.

The same support page says updates roll out gradually, depend on the carrier and device, and can take a few weeks to arrive. A phone bought later in its sales life does not receive a new seven-year clock.

APPI News’s phone-update guide separates security support from app compatibility, backup preparation and short-term problems after an operating-system release. The length of support is one purchase factor; rollout timing and the ability to recover data remain separate checks.

The documented differences point to four distinct choices

The standard Pixel 11 is the lowest-cost option and retains Tensor G6, 12GB of RAM, a 5x optical telephoto camera and 25W Qi2.2 wireless charging. The smaller Pro keeps a similar body size but adds the brighter display, upgraded rear cameras, HiLight notifications and a path to 16GB of RAM through a higher storage tier.

The Pro XL carries the same Pro camera system in a larger, heavier body with a larger battery and faster wired-charging claim. The Pro Fold is the choice defined by its 204 mm inner display and folding design; it has 16GB of RAM but returns to a 30x maximum zoom.

A complete comparison should use the Taiwan price of the exact storage tier, not the family’s starting price. It should also treat Suica as unconfirmed unless Google or East Japan Railway Company publishes support for Taiwan-sold Pixel 11 hardware.

Frequently asked questions

How much does the Pixel 11 cost in Taiwan?
The standard Pixel 11 starts at NT$29,990, about US$941 at Taiwan’s August 24 closing rate. The Pro, Pro XL and Pro Fold start at about US$1,161, US$1,350 and US$1,915, respectively.

Do Taiwan-sold Pixel 11 phones officially support Suica?
APPI News could not verify that claim in Google or East Japan Railway Company documentation as of August 25, 2026. Google’s Taiwan Wallet table lists iPASS locally and routes Suica guidance to Japan-only help pages.

Does the 256GB Pixel 11 Pro have 12GB or 16GB of RAM?
Google lists 12GB for the 256GB Pixel 11 Pro and Pro XL. Their 512GB and 1TB versions have 16GB, while the Pro Fold has 16GB at every listed storage capacity.

When does the seven-year update period begin?
Google starts the clock when the model first becomes available on its US Google Store, not on an individual buyer’s purchase date. Updates can reach carriers and devices on different schedules.

Hospitals need independent tests for AI trained on synthetic data

APPI News Editorial — Sun, 23 Aug 2026 17:15:52 GMT

The International Electrotechnical Commission (IEC) published PAS 63621:2026 on March 18, 2026, setting out data-management processes for information used to train, test or validate artificial intelligence (AI) models in medical devices. The specification covers traceability, annotation, representation, privacy, versioning and the data lifecycle, but it stays at a high level and does not set a universal passing score.

The International Medical Device Regulators Forum (IMDRF) published 10 final good machine learning practice principles in January 2025 that call for representative clinical data, independent training and test sets, human-AI evaluation and monitoring after deployment. Together, the two documents place synthetic-data checks inside a larger evidence chain rather than treating statistical similarity as proof that a system is ready for clinical use.

Synthetic resemblance is only the starting point

Synthetic medical data are generated partly or entirely by computational methods instead of being acquired directly from a person. The US Food and Drug Administration (FDA) says its regulatory-science program is studying synthetic data as a way to supplement patient datasets that can be difficult to obtain because of cost, safety limits, privacy restrictions or low disease prevalence. The agency describes a research program, not a clearance route for synthetic datasets.

A generator can reproduce broad distributions while missing the rare feature that matters for a clinical task. A 2024 radiology review by nine FDA researchers distinguishes statistical fidelity from downstream utility and says primary research supporting device safety and effectiveness would continue to come from a robust test set of real patients. The paper is limited to radiological imaging and is not binding US FDA guidance.

A separate APPI News explainer covers how synthetic medical data are generated and why fidelity, privacy and recursive training need different tests. The broader implementation framework covers intended use, evidence, interoperability, human control and monitoring for medical AI. The five checks below connect those two stages for hospital review.

Five checks connect generated data to clinical evidence

Check 1: Record provenance and population coverage

A validation file first needs to identify the records, simulations or knowledge models behind the generator. It can name the source populations, care settings, equipment, collection dates, inclusion rules, labels, missing-data treatment, generator version and transformations applied after generation. Separate entries can identify groups, conditions or acquisition settings that the source did not cover.

The IEC specification's public summary includes dataset classification, annotations, traceability, metadata, representativeness and validity periods in its data-lifecycle scope. The IMDRF principles similarly call for training, testing and monitoring datasets that represent the intended patient population, use environment and measurement inputs. A large generated dataset cannot add evidence for a group that was absent from its source without an independently supported method for creating and testing that group.

Check 2: Isolate training from the clinical test

Independence has to be checked at the patient, site and acquisition level. A random row split can still leak information when several records belong to one patient, when synthetic examples derive from people in the test cohort or when the same hospital and device settings dominate both sides. The provenance record needs to follow the original data into the generator, the AI training set, model tuning and the final test set.

The IMDRF principles say training and test datasets should remain appropriately independent and that dependence related to patients, sites and data acquisition must be addressed. They also say the extent of external validation should be proportionate to risk. For radiology, the FDA-authored review says a robust real-patient test set would remain the primary evidence for safety and effectiveness even when synthetic data support development or supplementary testing.

Check 3: Match the metrics to the clinical task

The US FDA says classification, estimation, image segmentation, time-to-event analysis and abnormality detection can require different performance measures, while label uncertainty and the way outputs are presented can change which measures are suitable. That rules out one accuracy score as a universal acceptance test.

The intended decision sets the test design. A classification review can report error types at the operating threshold and results for prespecified groups, while an image-segmentation review needs a reference standard that matches the clinical use. A text system needs tests for whether its output can be checked in the available workflow and whether unsupported content is detected before it reaches a decision.

Synthetic-data fidelity, downstream utility, clinical performance and privacy answer separate questions. Similar distributions do not show that a model performs on new patients, and good task performance does not establish that the generated records protect their sources. Each claim needs its own method, test population and uncertainty estimate.

Check 4: Test the interface and the human-AI workflow

Joint principles from the US FDA, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency say device information should cover intended users, populations, inputs, outputs and workflow effects, and they list local acceptance testing among good practices. They also call for known data gaps, failure modes and model changes to be communicated when users need that information.

A hospital test can follow a result from patient selection through data retrieval, model output, human review and the final record. It can measure whether identifiers, units, timestamps and missing values arrive intact; whether latency fits the use; whether users see limitations; and whether an unusable result follows a defined fallback path. This assesses the human-AI team rather than the model in isolation.

Health Level Seven International (HL7) defines Fast Healthcare Interoperability Resources (FHIR) as a standard for electronic health-information exchange built from modular Resources that can be combined for specific uses. A successful FHIR exchange does not by itself show that local codes, units, permissions or clinical context were mapped correctly, and it says nothing about the model's clinical performance. Those require interface and clinical acceptance tests.

Check 5: Monitor drift and control model changes

The IMDRF's final principle calls for ongoing monitoring in real-world use and controls for retraining risks such as overfitting, unintended bias and degradation caused by dataset drift. A monitoring record can identify the deployed model and data versions, baseline performance, review schedule, alert thresholds, investigation owner, stop condition and rollback version. The same record needs to distinguish a data-pipeline change from a model update.

The World Health Organization's 2021 evidence framework is intended to guide developers, researchers, policy-makers and implementers from AI medical-device development through post-market surveillance. It does not establish one monitoring interval or change rule for every country. Regulatory status, reporting duties and authorization for product changes still have to be checked in each market.

The evidence package has to tell one consistent story

The five-check sequence is an APPI News synthesis, not an IEC or IMDRF certification scheme. A procurement file can turn it into a connected set of deliverables: the intended-use statement, source-data and generator records, an independent clinical evaluation, task-specific results, interface mappings, a workflow test, and monitoring and change-control plans. Each item needs to describe the same population, inputs, output, setting and model version.

This separation gives reviewers several stopping points. A dataset can fail its provenance review before integration work begins; a model can pass a benchmark but fail on an independent site; and a valid model can still be held back by an interface that loses units or a workflow with no fallback. Passing one check cannot compensate for missing evidence at another.

The cited documents do not provide a universal numeric threshold for synthetic-data quality. APPI News also could not find a comparable international dataset showing how many hospitals require all five checks in routine procurement at the time of writing. The framework therefore describes an evidence structure, not current adoption or proof that a particular system is safe, effective or authorized in any country.

Synthetic data can reduce some barriers to model development, rare-case simulation and standardized testing. It does not remove the need to establish provenance, test on independent clinical evidence, verify the real workflow and watch the deployed version for change. A hospital can accept synthetic data for a defined role without treating the dataset as a substitute for the rest of the clinical evidence.

Phone RAM choices turn on app reloads and on-device AI support

APPI News Editorial — Sun, 23 Aug 2026 01:05:26 GMT

Google updated its Play for On-device AI guide on August 14, 2026, documenting model-delivery rules that can target phones by installed random-access memory (RAM). The rules do not create a universal 16GB threshold, while Android and iOS memory documents point to background app retention as the more immediate difference.

The official documents reviewed for this guide do not designate 12GB as sufficient or 16GB as necessary. They explain operating-system behavior and developer controls, not a controlled comparison of otherwise identical phones. The source report's four-phone result is omitted because APPI News could not verify its raw data, firmware versions or full test procedure.

RAM differences are most useful when measured against a repeatable set of everyday apps.

Background reloads are the first difference to test

Android keeps non-foreground app processes in a cache so it can reuse them when a user returns, but removes cached processes as memory runs low. A retained process can resume more quickly, while a removed process has to reload. The documentation does not promise that a given RAM capacity will retain a fixed number of apps.

Google says Android's low-memory killer removes background apps before foreground and system processes, and device manufacturers can change the mechanism's behavior. Two phones with the same advertised RAM can therefore manage the same app sequence differently. The memory used by each app also changes the point at which the system comes under pressure.

Apple says iOS can move a memory-intensive background app to solid-state storage, adding delay when it returns to the foreground, and that an app's memory limit depends on the device. Android and iOS implement memory management differently, but neither capacity label gives a guaranteed count of apps that will stay ready.

Operating-system policy affects which background apps remain available for a quick return.

More RAM is capacity, not a general speed rating

The official pages describe memory pressure, caching and process termination. They do not provide evidence that changing an otherwise identical phone from 12GB to 16GB raises game frame rates or shortens the cold launch of one app. Either claim requires a controlled test of the exact models and software involved.

A repeatable workload is more useful than a generic count of opened apps. On both candidate phones, the comparison can open the same game or editing project, browser tabs, camera and messaging apps, switch through them in the same order, and then record which apps retain their state. Matching software versions and settings are necessary so the test measures more than a RAM label.

Phones with the same RAM capacity can behave differently because their software and workloads differ.

AI support must be checked on the exact phone

Google's Android AI overview says on-device AI uses local storage, RAM and battery, requires compatible Android hardware, and can vary with operating-system and hardware support. Installed RAM is therefore one input, not an eligibility list. A 16GB specification cannot show whether a particular feature runs locally, uses a cloud model or is supported on that phone.

Google Play's beta delivery guide lets developers target model variants by device RAM, gives devices with at least 6GB as one example, and caps each AI pack at 1.5GB of compressed download size. The 6GB figure is a distribution example rather than a general runtime minimum, while 1.5GB describes a download package rather than RAM use. The exact phone's product page and feature-support list still have to establish compatibility.

A larger RAM figure does not establish that a named AI feature runs on the device.

Extra memory does not extend software support

RAM capacity does not set the period for operating-system or security updates. APPI News's phone-update guide separates a device's support status, release notes and required-app compatibility from its other hardware specifications. Those checks belong beside RAM when the phone is intended for long-term use.

Paying for 16GB buys usable margin only when the phone's workload or a supported local model can use it. It does not prove that future AI features will arrive, that the manufacturer will extend updates or that an aging battery will remain serviceable. Vendor commitments and local repair terms need their own evidence.

RAM capacity and the manufacturer's software-support period are separate purchasing checks.

A conditional rule for choosing 12GB or 16GB

The 12GB option can be the stronger value when it costs less, the exact model supports the required features and a repeatable workload does not force needed apps to reload. That is a conditional purchasing judgment, not a claim that 12GB covers every user, app set or phone.

The 16GB option has a clearer use when a matched test shows disruptive reloads on the 12GB model, or when official documentation for an intended on-device feature lists a higher memory requirement. An “AI phone” label by itself does not meet the second condition.

If neither distinction appears in testing or official support information, the extra 4GB has no verified benefit for that buyer's current use. Price, storage capacity, software-support period and repair terms can then decide which phone offers more measurable value.

The useful distinction is whether extra RAM changes an identified workload or feature requirement.

Frequently asked questions

Is 12GB enough for ordinary phone use?
The official documents reviewed here do not set a universal threshold. A 12GB model is the defensible choice when it supports the required features and retains the apps used in a repeatable real-world test.

Will 16GB improve gaming frame rates?
Not on the evidence cited in this guide. The official documentation explains process retention and memory pressure but provides no 12GB-versus-16GB frame-rate benchmark, so that claim requires testing of the exact phone and game.

Does on-device AI require 16GB?
Google publishes no universal 16GB requirement in the documents reviewed here. Its tools let developers target models by RAM, while actual availability also depends on the device, operating system and model deployment.

How can two RAM options be compared before purchase?
Run the same demanding apps in the same order on both versions and note which ones reload or lose state. Confirm the exact model's update commitment and AI support separately, because neither is guaranteed by installed RAM.

Metformin study highlights the limits of medical AI explanations

APPI News Editorial — Sun, 23 Aug 2026 00:00:00 GMT

A 2025 Science Advances study reported that low-dose metformin's glucose-reducing effect in mice depended on Ras-related protein 1 (Rap1) in the brain. The experiments did not test whether the pathway has the same importance in people receiving routine oral treatment. They instead add one candidate mechanism to a drug whose biological action has remained the subject of continuing research.

The paper matters less as a verdict on metformin than as an example of how evidence accumulates. A measured outcome, a proposed mechanism, and an applicable patient population are separate claims, and each requires a different test.

The mouse study identifies one pathway

The researchers compared control mice with animals that lacked Rap1 in the forebrain. The knockout mice had a weaker glucose response to low-dose metformin but remained responsive to other glucose-reducing agents tested in the study. Experiments that changed Rap1 activity and measured neurons in the ventromedial hypothalamus gave the proposed pathway biological support.

The design also limits what the result can establish. It used gene-altered mice and high-fat diets, and some experiments administered metformin directly to the brain. Those procedures cannot determine whether the same pathway drives long-term oral treatment in people.

An explanation and a mechanism are not equivalent

A drug-mechanism experiment asks whether changing a biological pathway changes an observed effect. An explanation tool for artificial intelligence may instead display a heat map, feature ranking, or other account of why a model produced an output. The comparison is useful for separating evidence layers, but the two forms of explanation use different methods and cannot be validated as if they were interchangeable.

A 2023 systematic review screened 882 papers on explainable AI in health care and included six; it found varied measures of fidelity, interpretability, plausibility, user satisfaction, trust, and task performance. The authors found no comprehensive, agreed framework for explanations or standardized method for evaluating whether they work for different users. A readable display therefore does not by itself establish that it faithfully represents the model's decision process.

Clinician trust can rise, fall, or stay unchanged

A 2024 systematic review screened 778 articles and included 10 studies that measured clinicians' trust: five reported an increase with explainable AI, three found no effect, and two found that explanations could increase or decrease trust. The reviewers rated every included paper at moderate or moderate-to-high risk of bias. Complex or contradictory explanations were among the factors that could reduce trust.

The review also separates trust from correctness. Greater trust in a wrong recommendation and lower trust in a correct recommendation can both reduce clinical accuracy. A trust survey cannot establish that a system improves patient care, and an explanation should be tested for whether it calibrates reliance on the model.

Validation has to follow the intended use

The FUTURE-AI consortium built its 2025 consensus guideline through work involving 117 experts from 50 countries and set 30 practices covering design, development, validation, regulation, deployment, and monitoring. Its six principles are fairness, universality, traceability, usability, robustness, and explainability. The practices span technical, clinical, socioethical, and legal questions rather than treating model accuracy as a complete evidence package.

For a medical AI claim, the evidence chain starts with a defined task, population, setting, and model version. Independent testing needs data that reflect sites, equipment, and records outside the development set. Prospective evaluation can then compare care with and without the system using predefined measures such as missed cases, false alerts, workload, and patient outcomes.

Deployment adds another layer. Developers and health organizations need records of version changes, data drift, incidents, cybersecurity controls, and human overrides. Monitoring those events tests whether performance persists after the model leaves its original dataset.

Global principles do not create a universal approval

The World Health Organization's 2023 publication grouped its considerations into documentation and transparency, life-cycle risk management, intended use and external validation, data quality, privacy and data protection, and stakeholder collaboration. The publication addresses governments and regulatory authorities across different settings. It supplies common questions, not an authorization for a particular product.

Whether software is regulated as a medical device depends on its intended use and the law in each market. Approval, permitted claims, and availability must therefore be checked by country. This report does not assess the status or deployment record of any named medical AI product.

What the metformin example does and does not prove

The mouse study adds a candidate biological explanation, but it does not settle metformin's mechanism in people. The AI reviews show that a plausible-looking explanation may affect clinician trust, while the small and inconsistent evidence base leaves its effect uncertain. Neither finding validates a medical AI system.

A medical AI case becomes stronger when performance replicates outside the development dataset, benefit is measured prospectively in the intended workflow, people can intervene, and post-deployment changes are tracked. The explanation remains one part of that record rather than a substitute for it.

Minor wounds need pressure, rinsing and a clean dressing

APPI News Editorial — Sat, 22 Aug 2026 22:50:04 GMT

England's National Health Service (NHS) last reviewed its cuts-and-grazes guidance on April 2, 2026, and lists three core steps for a small wound: control bleeding, rinse with water and apply a clean dressing. That sequence is meant for a superficial injury with no embedded object and bleeding that can be controlled.

Healthdirect Australia likewise advises firm pressure, clean water or saline and a non-stick dressing for a minor wound. Deep, gaping, contaminated or bite wounds need medical assessment, while severe or spurting bleeding calls for the local emergency service.

Severe bleeding and hidden damage need urgent care

Healthdirect Australia tells people to call an ambulance for severe or spurting bleeding, faintness, confusion, shortness of breath or loss of consciousness. Keep firm pressure on the wound while help is arranged, unless an embedded object makes direct pressure unsafe.

MedlinePlus, a service of the US National Library of Medicine, says severe bleeding that has not stopped after 10 minutes of direct pressure requires the local emergency number. Do not use that timing to delay help when bleeding is already heavy, spurting or accompanied by reduced consciousness.

England's NHS directs people to emergency care for very large or deep wounds, serious cuts on the face or palm, loss of feeling, impaired movement or an embedded object such as glass. Do not pull out an embedded object; press on either side of it without pressing directly on the object.

Seek prompt medical assessment for a human or animal bite, a deep puncture, a gaping wound, a wound that remains dirty after rinsing or an uncertain tetanus vaccination history. Healthdirect also lists increasing pain, redness, swelling, warmth, pus, fever or failure to improve after about five days as warning signs that need clinical review.

Clean a minor wound in three stages

Look for increasing redness, swelling, warmth or pain, pus, fever or a red streak extending from the wound whenever the dressing is changed. MedlinePlus lists those changes as signs that require prompt medical contact.

Hydrogen peroxide and antiseptic creams are not routine steps

England's NHS home-care sequence uses water, drying and a clean dressing, while Healthdirect Australia says antiseptic creams are not needed for routine minor-wound care. Neither sequence makes alcohol, iodine or hydrogen peroxide a standard step.

The AAD advises cleaning with soap and water and warns that hydrogen peroxide can further damage skin. This does not set the treatment for bites, punctures, surgical wounds, chronic wounds or wounds already under clinical care; a clinician may choose a different cleaning method, closure or medicine for those injuries.

Official consumer guidance is not uniform on topical antibiotic ointment. MedlinePlus includes antibacterial ointment in its minor-cut steps, while the AAD tells readers not to apply topical antibiotics; this guide therefore makes no topical product recommendation.

Bites, punctures and contaminated wounds need a tetanus check

The US Centers for Disease Control and Prevention's June 2025 guidance classifies penetrating or puncture wounds and wounds containing dirt, soil, feces or saliva, including animal and human bites, as dirty or major wounds. It tells clinicians to assess the wound and vaccination history before deciding whether tetanus vaccination or tetanus immune globulin is indicated.

The CDC document is US clinical guidance, not a universal vaccination timetable. A clinician should apply the tetanus schedule and care pathway used in the country where treatment is provided, especially when the vaccination history is unknown or incomplete.

Healthdirect Australia advises medical review for any human or animal bite and for a wound that is dirty, deep or difficult to clean. Washing remains a first-aid step, but it cannot settle the need for vaccination, wound closure or other clinical treatment.

The World Health Organization calls for early medical management after an animal bite and advises washing a dog-bite wound with soap and running water for 15 minutes. That 15-minute instruction is specific to dog bites and should not be treated as a universal rinsing time for ordinary cuts.

Children, pregnancy, diabetes and regular medicines change the threshold

Children

Healthdirect applies its minor-wound sequence to adults and children and uses the same referral signs for a child whose wound will not stop bleeding, is dirty, is deep or follows a bite. Those injuries need clinical assessment rather than an extended attempt at home cleaning.

Pregnancy

APPI News could not find a separate pregnancy-specific minor-wound cleaning sequence in the cited national guidance at the time of writing. A pregnant person considering a pain medicine, ointment or antiseptic should ask a clinician or pharmacist, while severe bleeding and the other urgent signs above still require immediate help.

Diabetes or severe immunodeficiency

Healthdirect advises people with diabetes to discuss wound care with a health professional because foot wounds can go unnoticed and may heal more slowly. The US CDC also makes severe immunodeficiency part of a clinician's tetanus immune globulin assessment for dirty or major wounds.

Blood-thinning medicines and other regular medicines

Healthdirect advises ambulance care when a deep wound continues to bleed under pressure in a person taking blood-thinning medicine. Take a current medicine list to urgent care, and do not stop or change a regular medicine as part of first aid unless a treating clinician directs it.

Common errors can make assessment harder

MedlinePlus warns against assuming that a wound is clean because no debris is visible, breathing on an open wound, probing for debris or removing a deeply embedded object. Rinse a minor wound, but leave deep material in place for medical assessment.

Frequently asked questions

What comes first, stopping the bleeding or rinsing?
Control active bleeding with firm pressure before rinsing a minor wound. Call the local emergency service if the bleeding is severe, spurting or will not stop.

Does every small cut need a disinfectant?
No routine antiseptic appears in the cited NHS or Healthdirect sequence for a small, clean wound. Water, gentle drying and a clean dressing are the basic steps, and the AAD warns against hydrogen peroxide.

Which wounds need a tetanus assessment?
Punctures, bites and wounds contaminated with soil, feces or saliva need clinical review, particularly when the vaccination history is unknown or incomplete. The clinician should use the schedule in the country where care is given.

When should a dressing be changed?
Replace it whenever it becomes wet, dirty or loose, and keep the wound covered while it remains open. Increasing pain, redness, warmth, swelling, pus or fever calls for medical assessment.

Chemical eye splashes require immediate rinsing before assessment

APPI News Editorial — Sat, 22 Aug 2026 22:49:25 GMT

The 2024 US first-aid guidelines from the American Heart Association (AHA) and American Red Cross call for immediate irrigation after chemical exposure to an eye and advise copious tap water for 15 minutes unless local guidance says otherwise. The guidelines also say contaminated water must be kept away from the other eye, nearby skin and other people.

England's National Health Service (NHS) and Healthdirect Australia both use a 20-minute target. Start the water before looking up ingredients, and continue rinsing while urgent help is arranged for a strong or unidentified product or when warning signs appear. The duration is a practical target, not a universal safe stopping point.

Strong chemicals and sight changes need emergency care

Call the local emergency medical service while rinsing after exposure to a strong chemical such as oven cleaner or bleach. England's NHS also lists any sight change, severe eye pain, headache or light sensitivity, nausea or vomiting, inability to move or keep the eye open, and blood or pus from the eye as reasons for immediate emergency care.

Healthdirect Australia takes a broader position, directing anyone with a chemical in the eye to an emergency department or ambulance and identifying worsening pain, blurred or lost vision, black spots, blood and colored discharge as urgent warning signs. That route applies in Australia; readers elsewhere should use the emergency service where they are located.

Do not drive when pain or impaired sight could make travel unsafe. Have another person retrieve the product container, label or workplace Safety Data Sheet (SDS) while rinsing continues, because identifying the substance must not delay irrigation.

Use gentle running water and protect the unaffected eye

The NHS says to use clean water that is not hot, hold the eye open, run plenty of water over it for at least 20 minutes and keep the flow from becoming too strong. A tap, shower or clean bottled water can provide the flow when no eyewash station is available.

Healthdirect advises tilting the head back and toward the injured side, holding the eyelids open and flushing with cool water for 20 minutes without letting runoff enter the unaffected eye. Keep the stream gentle and directed across the eye rather than using a high-pressure jet against the eyeball.

Canada's national occupational health and safety center, the Canadian Centre for Occupational Health and Safety (CCOHS), says water flushing must begin immediately and that Section 4 of a hazardous product's SDS should contain product-specific first-aid instructions. Another person can locate those instructions after irrigation has started.

Contact-lens removal comes after the water starts

CCOHS says to start flushing immediately, remove contact lenses as soon as possible and never delay irrigation while trying to take them out. A lens that releases easily can come out during rinsing, followed by continued water flow.

Healthdirect advises trying to remove contact lenses after chemical exposure. If a lens does not release without force, continue irrigating and tell the responding clinician rather than repeatedly pinching or pulling at the eye.

Rubbing and household neutralizers can add injury

The NHS says not to touch or rub an injured eye, and Healthdirect says not to put unprescribed drops, ointments or medicines into it. Rubbing, wiping with a towel or probing under the eyelids does not replace continuous irrigation.

CCOHS warns that trying to neutralize an acid with a base, or a base with an acid, can delay irrigation, generate heat during the reaction and expose the eye to another damaging substance. Do not add vinegar, baking soda or another household chemical; begin with water and follow verified product instructions or clinical direction.

First aid for skin burns follows a different sequence and should not be used as a substitute for open-eye irrigation. APPI News's burn first-aid guide separates thermal, chemical and electrical injuries, while a chemical eye splash follows the eye-specific steps in this guide.

Children and other higher-risk groups need separate judgment

Children

Australia's Royal Children's Hospital Melbourne says strong acid or alkali eye exposures in children require immediate copious irrigation and that every chemical burn should be discussed with an ophthalmologist. A child may need adult help to keep the eyelids open and maintain a gentle stream. If irrigation cannot continue safely, obtain emergency help; England's NHS also sets a lower threshold for urgent advice when a child has an eye injury.

Pregnancy

APPI News could not find a separate initial irrigation method for pregnancy in the cited guidance at the time of writing. Begin water irrigation immediately, then tell the treating clinician about the pregnancy so any further examination or medicine decision can be individualized.

Chronic illness and regular medicines

The cited guidance does not publish a different initial home-rinsing sequence for chronic illness or regular medicine use. Take a current medicine list or the medicine containers to clinical assessment, and leave decisions about eye drops or other treatment to the clinician.

Twenty minutes is a target, not a universal stopping rule

The 2024 US AHA and American Red Cross guideline uses 15 minutes unless local guidance says otherwise, while the NHS and Healthdirect use 20 minutes. The US guideline says studies comparing different irrigation durations are not available and bases its 15-minute recommendation on an observational study and an animal experiment.

The 20-minute figure is therefore a practical minimum from the cited English and Australian public guidance, not proof that every chemical has been cleared at that point. Continue rinsing when an emergency dispatcher, poison specialist, product label or SDS calls for longer, and do not cancel medical assessment because pain briefly eases.

Frequently asked questions

What should happen first after a cleaning product reaches an eye?
Begin a gentle flow of clean, cool or lukewarm water immediately and hold the eyelids open. Product identification, contact-lens removal and calls for help can proceed without stopping the rinse.

Can vinegar or baking soda neutralize the product?
No household acid or base should be added to the eye. CCOHS warns that neutralization can delay water, release heat and cause injury from the second substance.

What if a contact lens will not come out?
Keep rinsing and stop repeated forceful attempts to remove it. Tell emergency or eye-care clinicians that the lens remains in place so they can assess the surface and remove it safely.

Does feeling better after 20 minutes rule out an injury?
No. Product strength, concentration and contact time vary, and symptom relief cannot establish that the eye surface is unharmed or that irrigation is complete.

Watery diarrhea care starts with rehydration and warning signs

APPI News Editorial — Sat, 22 Aug 2026 22:41:18 GMT

The World Health Organization (WHO) identifies dehydration as the most severe immediate threat from diarrhea. WHO says liquid stools remove water and electrolytes and lists oral rehydration salts (ORS), continued feeding and clinical review for blood in the stool, dehydration or persistent diarrhea among the core responses.

A person who remains alert, can drink and has none of the warning signs below can begin replacing losses while monitoring urination, vomiting, stool and alertness. Watery stool alone does not identify the cause. APPI News separately explains why suspected food poisoning requires attention to dehydration, bloody diarrhea and repeated vomiting rather than assumptions about the last meal.

Altered alertness, blood and inability to drink need urgent care

Confusion or failure to respond normally, severe breathing difficulty and sudden severe abdominal pain call for local emergency medical help. England’s National Health Service (NHS) directs people to emergency care for those signs. Emergency numbers and routes into care differ by country.

The US National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) advises contacting a doctor right away for altered mental state, frequent vomiting, severe abdominal or rectal pain, black or tarry stool, red blood or pus in stool, or signs of dehydration. Extreme thirst, dry mouth, reduced or dark urine, dizziness and sunken eyes or cheeks are among the dehydration signs it lists. Blood or black stool requires prompt assessment, although the route into care depends on the person’s condition and local health system.

The NHS also advises urgent assessment when vomiting prevents liquids from staying down, diarrhea is bloody or dehydration signs continue after ORS use. Repeated attempts to drink at home should not delay care when no fluid stays down. WHO says severe dehydration or shock may require intravenous fluids.

The US NIDDK adds time and frequency thresholds: adults should contact a doctor for diarrhea lasting more than two days, six or more loose stools in one day or high fever. Its guidance uses a threshold of more than one day for children and calls for prompt advice for any fever in an infant or refusal to eat or drink for more than a few hours. These are US public-health thresholds, not verified worldwide cutoffs, and no clock should delay care when a warning sign is already present.

Oral rehydration replaces both fluid and electrolytes

A 2006 WHO and UNICEF standard says an adequate glucose-electrolyte ORS can treat dehydration from diarrhea in all age groups except the most severe cases. The document is a pharmaceutical manufacturing standard, not a kitchen recipe. This guide therefore does not reproduce ingredient weights or set a fluid volume.

Packaged ORS should be prepared with the amount of safe water stated on its label, without adding extra powder or changing the concentration. The NHS advises taking small sips when nausea makes drinking difficult. If even small amounts will not stay down, urgent clinical assessment is warranted.

A short written record can include how often watery stools and vomiting occur, whether liquids stay down, how urination changes and whether alertness declines. Those observations can show deterioration, but they cannot establish the cause or measure dehydration precisely.

Eating can resume as appetite and tolerance return

The US NIDDK says adults can usually return to their normal diet when appetite returns, children should continue an age-appropriate diet, and infants should receive breast milk or formula. Its guidance does not recommend fasting or a broadly restricted diet for acute diarrhea.

The same NIDDK guidance says alcohol, caffeinated drinks, foods and drinks high in simple sugars, sugar alcohols and high-fat foods can worsen acute diarrhea for some people. A temporary loss of appetite is not a reason to impose a prolonged clear-liquid diet.

The NHS says breast or bottle feeding should continue and infant formula should remain at its usual strength rather than being diluted. Smaller, more frequent feeds may be used when a baby is vomiting, but poor feeding, fewer wet diapers or unusual drowsiness requires prompt pediatric advice.

Symptom medicines are not a universal first step

The US NIDDK says doctors typically advise against over-the-counter antidiarrheal medicines for infants, children, or people with blood in the stool or fever. That guidance does not create a blanket ban for every adult episode. A clinician or pharmacist can assess whether a symptom medicine is appropriate for the person, other medicines and the country where it would be used.

Medication should not replace rehydration or delay assessment of a warning sign. APPI News could not verify one instruction for stopping, continuing or changing every regular medicine during diarrhea. People taking routine medicines should follow any existing clinician-issued sick-day plan and contact the prescribing clinician or a pharmacist rather than infer a dose change from a general guide.

Children, pregnancy and existing illness change the threshold

Infants and children

The US NIDDK advises prompt medical help when a child with diarrhea is younger than 12 months, was born prematurely, has another medical condition or cannot drink enough liquid or ORS to prevent dehydration. Adult fluid instructions should not be applied to a child. The NIDDK advises caregivers to ask a doctor before giving ORS to an infant.

Pregnancy

The US NIDDK lists pregnant people among those more likely to develop health problems from diarrhea and advises them to keep in touch with a doctor. Vomiting that prevents drinking, reduced urination, blood in the stool or another warning sign warrants prompt assessment. APPI News could not verify a separate international home-ORS protocol for uncomplicated watery diarrhea during pregnancy.

Older adults, weakened immunity and chronic illness

The US NIDDK advises older adults and people with weakened immunity, diabetes, kidney disease or another health condition to talk with a doctor before using ORS. The appropriate solution or fluid amount may differ with the underlying condition, so a general volume is not supplied here.

Regular medicines

The US NIDDK identifies current antibiotic use as a factor associated with a higher risk of problems during diarrhea and lists medicine side effects among the common causes of acute diarrhea. APPI News could not verify one international rule for adjusting regular medicines during a diarrheal illness. A prescriber or pharmacist can assess a specific medicine without turning a general guide into a dosing instruction.

Frequently asked questions

Is plain water enough for watery diarrhea?
Water contributes to fluid intake, but ORS contains glucose and electrolytes designed to replace diarrheal losses. Signs of dehydration, substantial ongoing losses or an inability to keep fluids down require clinical advice rather than reliance on one drink.

How long can watery diarrhea be watched at home?
The US NIDDK advises adults to contact a doctor after more than two days or six or more loose stools in one day, and it uses a threshold of more than one day for children. Blood or black stool, severe pain, frequent vomiting, altered alertness or dehydration signs require earlier assessment.

Does eating need to stop until diarrhea ends?
No general fasting period is recommended in the cited guidance. Food can return with appetite and tolerance, while children continue an age-appropriate diet and infants continue breast milk or full-strength formula.

Sudden hip pain needs a weight-bearing check before home care

APPI News Editorial — Sat, 22 Aug 2026 22:28:23 GMT

England's National Health Service (NHS) says severe hip pain after a fall, inability to walk or bear weight, and tingling or loss of sensation after an injury require emergency care. Its adult hip-pain page was last reviewed on November 20, 2025.

Hip pain can arise from the joint, nearby soft tissue or pain referred from the back. The first decision therefore turns on the injury, weight-bearing ability and warning signs rather than an assumed diagnosis.

Loss of weight bearing or injury signs require urgent care

Call the local emergency service or seek emergency care for severe hip pain after a fall or other injury, inability to walk or put weight on the leg, or tingling or loss of sensation after an injury. The NHS also advises urgent assessment for severe pain that starts suddenly without an injury, a hot or swollen hip, skin color change, or hip pain accompanied by fever, chills or feeling unwell.

MedlinePlus, a service of the US National Library of Medicine, adds leg deformity, marked bruising or bleeding to its emergency list and calls for emergency help when worsening hip pain occurs with difficulty urinating or passing stool. These signs require assessment; they do not identify the cause by themselves.

Healthdirect Australia uses a broader prompt-review threshold: sudden or intense pain, pain after a fall or injury, inability to move the hip or leg, inability to bear weight, swelling, bleeding or fever. This differs from the NHS and MedlinePlus emergency wording, so mild pain after a fall should not be declared safe for home observation from symptoms alone.

Home measures come after the urgent screen

Home measures are an option only after the urgent signs above have been checked. Because loss of weight bearing appears on the urgent-care lists, an ice pack or sleeping-position change should not delay assessment when walking or bearing weight is unsafe.

For adults outside its urgent thresholds, the NHS advises avoiding excessive strain while continuing to move if able, without overdoing activity, and avoiding heavy loads. It also advises wrapping an ice pack or bag of frozen vegetables in a towel and applying it to the painful area for no more than 20 minutes every two to three hours.

Healthdirect Australia lists rest, heat or ice and gentle exercise among possible home measures. It does not set an ice duration or say that cold must always come before heat, and its prompt-review advice after a fall still takes priority.

MedlinePlus advises avoiding activities that worsen the pain and sleeping on the painless side with a pillow between the legs. It does not give an ice schedule on its hip-pain page.

The sources do not set one follow-up deadline

MedlinePlus advises contacting a clinician when hip pain remains after one week of home care, or when limping or difficulty with stairs or walking begins. The NHS uses a two-week threshold, but advises earlier review when pain worsens, returns, affects sleep or stops normal activities.

APPI News could not find a single international guideline that sets one follow-up timetable for acute hip pain at the time of writing. Worsening function or a new urgent sign overrides either country-specific waiting period.

Pain location does not establish the cause

Healthdirect Australia says hip pain may be felt inside the hip, groin, outer hip, upper thigh or buttock, and pain from the lower back can be referred to the hip. A location chart cannot replace an examination or determine whether home measures are appropriate.

MedlinePlus says clinicians assess where and when the pain started, what worsens it, walking and weight-bearing ability, other medical conditions and current medicines. A concise history can therefore cover onset, any fall, changes in walking or weight bearing, other conditions and a current medicine list.

Children, pregnancy and long-term steroids need separate decisions

Children

England's NHS says hip pain in a child should always be checked and calls for urgent assessment when a child develops sudden pain in the hip, thigh or knee, limps or cannot bear weight on one leg. Adult ice and activity advice should not be substituted for that assessment.

The same NHS page describes rest and reduced high-strain activity only after a child has been assessed as having irritable hip. It advises another clinical review if pain worsens or fails to improve, fever or chills develop, pain lasts two weeks, or the pain returns.

Pregnancy

England's NHS says a pregnant person should check with a pharmacist, midwife or doctor before taking any medicine, including a painkiller, and should not stop a prescribed medicine without first speaking to the prescriber. That page addresses medicine safety rather than the cause or treatment of hip pain.

Chronic conditions and regular medicines

MedlinePlus advises contacting a clinician when sudden hip pain occurs during long-term steroid use. The five cited pages do not provide a single hip-pain medication rule for every chronic condition or regular medicine, so people in these groups should discuss nonprescription pain medicine with a clinician or pharmacist.

Frequently asked questions

Does sudden hip pain mean ice should come first?
No single first treatment applies to every cause. Injury, weight-bearing ability, fever, swelling, skin changes and sensory symptoms come before home measures; the NHS ice schedule applies only after urgent care is not indicated.

Is walking after a fall proof that the injury is minor?
No. Healthdirect Australia advises prompt medical review for hip pain after a fall or injury, while the NHS and MedlinePlus use emergency wording for severe pain, loss of weight bearing or other serious signs.

How long can hip pain be watched at home?
The sources give different limits: one week in MedlinePlus and two weeks in the NHS adult guidance. Pain that worsens, returns, changes walking, affects sleep or normal activity, or develops an urgent sign needs earlier assessment.

Can heat be used instead of ice?
Healthdirect Australia lists both heat and ice as possible home options, while the NHS gives a timed, towel-wrapped ice method. These pages do not establish that one must always come first for every episode of sudden hip pain.

Taiwan exporters need plant data for EU importers' CBAM filings

APPI News Editorial — Sat, 22 Aug 2026 17:11:21 GMT

The European Commission published definitive-period guidance for operators outside the European Union on August 14, 2026. The release followed the January 1 start of the definitive phase of the Carbon Border Adjustment Mechanism (CBAM), under which authorized EU-side declarants file annual emissions declarations and surrender certificates.

For Taiwan manufacturers, the EU filing duty remains with the authorized CBAM declarant, usually the EU importer or sometimes an indirect customs representative. European Commission guidance says producers outside the EU are not themselves required to monitor, verify or register under CBAM, but they can provide verified actual emissions data; otherwise the declarant can use published default values.

The EU-side declarant carries the filing duty

The definitive CBAM regime took effect on January 1, 2026 and covers selected products in cement, iron and steel, aluminum, fertilizers, electricity and hydrogen. It applies a carbon cost to emissions embedded during production of covered imports as the European Union phases in the mechanism alongside changes to free allocation under the EU Emissions Trading System.

The legal responsibility sits on the import side. The consolidated EU regulation makes the authorized CBAM declarant responsible for the annual declaration and certificate surrender; this is normally an EU-established importer or, in defined cases, an indirect customs representative. A declarant can delegate submission, but remains responsible for compliance.

The 50-metric-ton rule is narrower than a general exemption for small exporters. For covered cement, iron and steel, aluminum and fertilizer goods, the threshold is the same importer's cumulative net mass across all covered Combined Nomenclature (CN) codes in a calendar year; once it exceeds 50 metric tons, that year's covered imports are subject to the obligations, including goods imported before the threshold was crossed. The mass-based exemption does not apply to electricity or hydrogen.

The first annual deadline is September 30, 2027

Each authorized CBAM declarant must submit a declaration by September 30 for the preceding calendar year, with the first filing due on September 30, 2027 for goods imported in 2026. The declaration records imported quantities and origins, embedded emissions, verification information and any eligible carbon price paid outside the European Union. The same date is the first deadline for surrendering the corresponding CBAM certificates.

Certificate purchases for 2026 imports begin in February 2027. The European Commission calculates four quarterly certificate prices for 2026 from weighted average EU emissions allowance auction prices, then moves to weekly prices from 2027. The price for each 2026 quarter applies to emissions in goods imported during that quarter.

A carbon price paid where the goods were produced does not create an automatic deduction. The EU regulation requires the declarant to document the carbon price effectively paid after rebates or other compensation and to obtain independent certification for the supporting evidence. APPI News did not verify whether any payment under Taiwan's carbon-pricing rules satisfies those conditions for a particular import.

Actual values begin at the producing installation

A company-wide greenhouse gas inventory does not by itself produce the figures required for every covered product. The European Commission's August guidance requires an installation operator using actual values to establish a monitoring plan, attribute emissions to production processes and compile an operator's emissions report for each eight-digit CN code. The data must cover product quantities, direct emissions, relevant indirect emissions and the embedded emissions of covered precursor materials.

Production routes cannot be selected only because one line reports lower emissions. The EU methodology groups routes where the rules require joint monitoring and assigns emissions within defined system boundaries. It also limits indirect-emissions reporting in the definitive phase to goods for which those emissions are within scope.

The operator's records must connect each number to a reporting period, method and source. The Commission guidance calls for the monitoring plan to be prepared in English and for monitoring records and supporting documents to be retained for at least six years. This makes meter records, fuel and electricity data, production volumes, precursor reports and calculation versions part of the evidence chain rather than a separate sustainability presentation.

Verified actual values and defaults are separate routes

EU rules allow annual declarations to use actual verified emissions or European Commission default values, while actual data must be checked by an independent verifier accredited by an EU national accreditation body. Verification takes place at the producing installation. The verifier reviews the monitoring approach, calculations and supporting evidence before issuing a report that the declarant can use.

Actual precursor data also require a verification report covering the relevant production period. If compliant actual data are not supplied, the declarant can use the applicable defaults. That choice is a permitted reporting route, not proof that the default represents the factory's measured performance.

Commission Implementing Regulation (EU) 2026/1740 corrected a limited set of definitive-period default values and applies from January 1, 2026. The act identifies transcription errors in default values for Taiwan steel products under CN codes 7218 through 7223. A supplier or importer using an earlier spreadsheet therefore needs to check the current legal act and version before completing a filing.

The Registry is an optional data-sharing route

The CBAM Registry's module for Operators of Third-Country Installations (O3CI) lets non-EU producers upload installation and emissions data once and disclose it to authorized declarants. A declarant can retrieve the shared data after providing its Economic Operators Registration and Identification (EORI) number to the supplier. The module is designed to keep business-sensitive material from being disclosed as part of the shared record.

Registration does not shift the filing duty to the producer. A registered operator can share the non-confidential part of verified data through the Registry; an unregistered operator can instead give the declarant the Commission-format emissions report and a verification report supplied through the verifier. The authorized declarant remains responsible for what enters the annual filing.

A working checklist for Taiwan suppliers

The operational task is to make the producer's records match the EU customer's customs and emissions reporting data. The following sequence reflects the European Commission's definitive-period guidance rather than Taiwan-specific reporting software.

Confirm which EU customer or indirect customs representative is the authorized CBAM declarant, and record the agreed document channel or Registry access route.
Match each covered product to the eight-digit CN code used for the EU import, the producing installation, the production process, the reporting period and the quantity produced.
Maintain an English monitoring plan that identifies system boundaries, calculation or measurement methods, data sources, product quantities and the controls used to find errors or missing data.
Separate direct emissions, indirect emissions where they are in scope, and emissions embedded in precursor materials. Record whether each precursor figure is verified actual data or a European Commission default.
Contract an accredited verifier early when actual values will be used, then retain the monitoring records, operator's emissions report and verification material for at least six years.
Document any carbon price claimed by the declarant as effectively paid outside the European Union, including rebates or compensation and the independent certification required by EU rules.
Version the default-value file, legal basis, formulas and data delivered to each customer so later corrections can be traced to the filing they affected.

The practical constraint is data alignment, not the submission of a Taiwan company-wide emissions total. The product code, installation, production records, precursor evidence and verification report must describe the same goods that the authorized declarant reports at the EU border. Commercial agreements can set timing and confidentiality, but they do not transfer the declarant's responsibility under EU law.

Coverage still requires a product-level check

CBAM coverage turns on the product's CN code, the producing installation and the EU importer's cumulative activity during the calendar year. APPI News did not classify any individual Taiwan exporter's goods, total an importer's shipments or determine whether a specific transaction qualifies for an exemption.

The evidence gap also extends to carbon-price deductions. APPI News verified the EU conditions but did not establish that a payment made under Taiwan's rules would satisfy them in a particular declaration. That omission is not a finding that a deduction is unavailable; it reflects the need for transaction records, independent certification and a decision under the applicable EU process.

Hospitals need five controls before putting medical AI to work

APPI News Editorial — Sat, 22 Aug 2026 17:11:21 GMT

The International Medical Device Regulators Forum (IMDRF) finalized 10 good machine learning practice principles for artificial intelligence (AI) in medical devices in January 2025. The document treats intended purpose, representative evidence, human-AI interaction and post-deployment monitoring as parts of one product lifecycle, leaving a model benchmark as only one piece of the deployment case.

Hospital teams can translate those principles into five implementation controls: purpose, evidence, interoperability, human control and monitoring after release. The sequence is an APPI News synthesis, not an IMDRF certification scheme, a regulatory approval or a legal test in any country.

Deployment begins with a decision, not an algorithm

A demonstration usually gives a model a prepared input and displays an output. Routine use adds patient identification, data retrieval, time pressure, incomplete records, staff interfaces, fallback procedures and a decision that may affect care.

A deployment specification therefore needs to identify the intended user, patient population, input, output and decision that the output may influence. It also needs to record when the system should not be used, what happens when it cannot produce a result and which person retains authority over the decision.

The World Health Organization's 2021 principles say humans should remain in control of health systems and medical decisions and call for accountability and redress when people are harmed by algorithm-based decisions. Those are governance requirements for the service around a model, not properties that can be established by its accuracy score.

Five controls turn a model into a clinical system

First control: Define the intended use before setting a test

A phrase such as “assists clinicians” is too broad to determine what evidence matters. A usable scope names the target population, intended users, input source and format, output, care setting, expected response time and exclusions. It also states whether the output prioritizes a worklist, flags a possible finding, drafts text or directly informs another type of decision.

Joint 2024 principles from the US Food and Drug Administration, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency say transparency should cover a device's medical purpose, target population, intended users, use environment, inputs, outputs and intended effect on decisions. The document presents good practices rather than a law that applies worldwide.

The scope determines the acceptance test. A system intended to rank cases for later review needs a different response to failure than one that produces information during an active consultation. If that difference is absent from the specification, a passing test cannot show that the model is ready for its stated workflow.

Second control: Test representative evidence under clinical conditions

The IMDRF principles call for clinical evaluation with data that represent the intended population, use environment and measurement inputs; they also call for independent training and test data and external validation proportionate to risk. Testing should examine the device in clinically relevant conditions and assess the human-AI interaction rather than the model in isolation.

Representation must follow the stated use. Evidence from one patient mix, scanner fleet or record system cannot by itself establish how a model performs after those conditions change. A hospital evaluation can prespecify important subgroups, missing-data conditions, equipment variations and workflow interruptions, then report where the available sample is too small to support a conclusion.

A headline accuracy figure can also hide different error patterns. The review record needs the measures that match the decision, results for relevant subgroups and a clear account of false positives, false negatives and unusable inputs. A result outside the validated boundary should trigger the fallback defined in the first control.

Third control: Validate data exchange, meaning and access

Health Level Seven International (HL7) defines Fast Healthcare Interoperability Resources (FHIR) as a standard for exchanging health information electronically, with modular Resources that can be combined and linked for specific uses. The standard gives systems a common structure for concepts such as patients, observations, diagnostic reports and medications. It does not make every source record complete or give local codes the same meaning.

Interface testing has to go beyond receiving a successful application programming interface (API) response. Patient identifiers, units, code systems, timestamps, missing values, source context and version identifiers need expected results that both technical and clinical teams can inspect. The implementation also needs to show that an output reaches the correct user without stripping away information needed to interpret it.

HL7 explicitly says FHIR is not a security protocol and that authentication, authorization and access-control decisions require a separate security system. The specification supplies resources that can support provenance and audit records, but an organization still has to define who may retrieve each record and how access is reviewed. A separate APPI News report examines how Taiwan is building a FHIR layer for hospital records and medical AI.

Fourth control: Put human authority into the workflow

The label “human in the loop” does not describe what a person can do. A working design defines whether a clinician may accept, modify, reject or mark an output unusable, how the system presents uncertainty and what path remains available during an outage. It also sets a route for reporting an error instead of treating every override as user resistance.

The joint transparency principles focus on the performance of the human-AI team and say information should appear when it is needed in the workflow, including notifications about updates. That means a static user manual is not enough when a changed model affects what appears on screen. Interface tests need to show that status, limitations and warnings are visible at the decision point.

An audit record can connect the model version, input time, output and subsequent human action. That record helps an organization reconstruct an incident and compare performance across versions, but it also contains sensitive operational or patient information that needs access and retention controls. The responsible teams and escalation route have to be named before the system enters routine use.

Fifth control: Monitor performance and control changes

The IMDRF principles treat deployment as the start of another evidence phase. They call for monitoring performance in real-world use and managing retraining risks such as unintended bias, overfitting and degradation caused by data drift. A monitoring plan therefore needs a baseline, review schedule, alert thresholds, investigation process, stop condition and rollback path.

Final US Food and Drug Administration guidance issued in August 2025 recommends that a predetermined change control plan describe planned modifications, the methods for developing, validating and implementing them, and an assessment of their effects. The agency reviews such plans within US marketing submissions for devices in the 510(k), De Novo and premarket approval pathways.

The US document is nonbinding guidance for specified US pathways, not a worldwide change rule. Its structure can still expose missing information in a hospital contract: which changes are permitted, which evidence is required, who authorizes release and how a previous version can be restored. Regulatory status and change requirements must be checked separately in every market where a device is used.

The evidence package has to connect the five controls

A pre-deployment file can combine the intended-use specification, data description, independent evaluation, interface mapping, clinical workflow, audit design, monitoring plan and change-control process. Each document should point back to the same population, inputs, outputs and decision. A mismatch, such as a test population narrower than the procurement claim, is evidence that the controls are not yet aligned.

Procurement scoring can separate model evidence, integration readiness and operational governance. A high benchmark score cannot compensate for an interface that loses units, a workflow with no fallback or a contract that permits unreviewed model changes. Conversely, a technically sound interface does not establish clinical performance.

The five controls also create separate decision points. A hospital can stop after the evidence review without spending on integration, pause an interface that fails semantic checks or roll back a changed version without discarding the original procurement record. That separation makes it possible to identify which claim failed instead of treating the entire project as a single pass-or-fail demonstration.

What the framework cannot establish

The cited documents are international principles, technical specifications and guidance from named regulators. They do not classify a particular product, authorize it for sale, set one evidence threshold for every clinical use or assign liability after an error. Those questions depend on the product, intended purpose, contracts and law in each country.

The five-control sequence is also not a published measure of hospital adoption. APPI News could not find a comparable international dataset showing how many hospitals use every control in routine medical AI deployments at the time of writing. Public documents often describe a framework, pilot or regulatory pathway without reporting whether a specific hospital has completed clinical validation and post-release monitoring.

Together, the cited documents define the shape of the evidence problem: intended purpose sets the boundary, representative testing checks performance, interoperability places the model in a real system, human control defines its role and monitoring follows changes after release. Within this five-control framework, the record is complete only when those documents describe the same system and the same clinical use.

Head injuries in older adults need checks before home observation

APPI News Editorial — Sat, 22 Aug 2026 15:34:09 GMT

The US Centers for Disease Control and Prevention (CDC) said in September 2025 that symptoms of a mild traumatic brain injury may appear immediately or hours or days after the injury. For an older adult, the first decision after a blow to the head is whether emergency or urgent assessment is needed; a fixed observation time cannot make that decision.

Home checks start only after a hospital or local health service has said home care is appropriate. The companion guide on recording alertness, walking, cognition and later changes after an older adult's head injury can organize information for clinicians, but it cannot rule out bleeding.

Emergency warning signs come before the clock

Call the local emergency service if the person loses consciousness, cannot be woken, has a seizure, develops weakness or numbness, or has trouble speaking, understanding, walking or coordinating movement. The US CDC also lists a worsening headache that does not go away, repeated vomiting, one pupil larger than the other, increasing confusion, agitation and unusual behavior as danger signs requiring immediate emergency care.

England's National Health Service (NHS) calls for emergency help after a head injury when clear fluid drains from the ears or nose, an ear bleeds, vision or hearing changes, or problems develop with balance, speech or comprehension. The service also directs people to urgent medical advice after vomiting or dizziness, or when the injured person has a bleeding disorder or takes medicine that reduces blood clotting.

Do not delay an emergency call to complete an observation list or find every medicine container. A caregiver can tell responders when the injury happened, whether consciousness was lost, how the person has changed and which medicines are known, then gather additional details without slowing transport.

Blood-thinning treatment lowers the threshold for assessment

The US CDC says traumatic brain injuries may be missed or misdiagnosed in older adults because symptoms can overlap with conditions such as dementia, and anticoagulant or antiplatelet medicines may increase the risk of bleeding in the brain after an injury. A familiar caregiver should describe the person's usual memory, speech, balance and independence so new changes are not mistaken for their baseline.

The United Kingdom's 2023 National Institute for Health and Care Excellence (NICE) guideline lists current anticoagulant treatment or antiplatelet treatment other than aspirin alone as a reason for emergency-department referral after a head injury. For people with no other scan indication, the guideline tells clinicians to consider computed tomography rather than treating a scan as automatic; this is a UK pathway, not a universal timetable.

A caregiver should not stop, skip or change a prescription medicine after the injury. The medicine names, the last known dose times and any anticoagulant alert card should go to the clinician who weighs the bleeding risk against the reason the medicines were prescribed.

Home care starts only after clinical advice

England's NHS says a person sent home after a minor head injury does not need to remain awake when tired, but an adult should stay with them for at least the first 24 hours. Any discharge instructions about waking, checks or return to care take priority over a generic schedule.

The accompanying adult should know the person's usual behavior and watch for changes in alertness, conversation, balance, limb use, headache, nausea and vomiting. A short written record can include the time and cause of the injury, any loss of consciousness or memory gap, each vomiting episode, symptom changes and a complete medicine list.

The 24-hour period is a supervision arrangement, not evidence that the risk has ended. The CDC says some mild brain-injury symptoms appear hours or days later, while England's NHS says symptoms of a subdural hematoma can begin immediately or develop gradually over several weeks.

Cold packs and rest do not replace monitoring

For a minor injury being managed at home, England's NHS advises applying a cold pack wrapped in cloth for short periods during the first few days. The guidance does not set a universal 15- or 20-minute interval, and a cold pack is not a reason to postpone care when another warning sign appears.

Rest is compatible with observation after a health service has approved home care. Emergency help is needed if the person becomes unusually drowsy, cannot be woken or develops another neurological warning sign; sleeping pills should not be added unless a doctor advises them.

Possible neck or spinal injury changes first aid

Severe neck or back pain, weakness, numbness or inability to move after a fall raises concern for a spinal injury. The Australian and New Zealand Committee on Resuscitation (ANZCOR) advises an alert person with a suspected spinal injury to remain as still as possible and says movement should be minimized if immediate danger makes relocation necessary.

Do not pull the person upright to test whether they can walk. Call the local emergency service and follow its instructions; airway and breathing take priority if the person becomes unresponsive.

Children, pregnancy and pre-existing conditions need separate decisions

Children

An older-adult checklist should not be applied to a child. The CDC adds inconsolable crying and refusal to nurse or eat to its child danger signs, while the NICE guideline uses separate assessment, scan and observation criteria for people under 16; a local pediatric service should set the next step.

Pregnancy

The neurological emergency signs remain reasons for immediate care during pregnancy, but a fall or collision can also create obstetric concerns. Pregnancy, Birth and Baby, an Australian government-funded health service, advises hospital assessment after even a minor injury when fetal movement decreases, vaginal bleeding occurs or abdominal pain develops.

APPI News could not find a separate evidence-based home observation schedule for an isolated head injury during pregnancy. A pregnant patient should use the head-injury return instructions and obtain obstetric advice if the fall involved the abdomen or is followed by bleeding, pain, contractions, fluid leakage or reduced fetal movement.

Dementia, bleeding disorders and regular medicines

Pre-existing cognitive impairment makes a reliable baseline especially useful, while a bleeding or clotting disorder requires prompt clinical advice after a head injury. The caregiver should provide all prescription and nonprescription medicines and avoid making medication changes without the treating clinician.

Frequently asked questions

Is the observation period 24 hours or several weeks?
There is no single period that makes every older adult safe. The first 24 hours describe close supervision after discharge or clinical advice, while new symptoms still require assessment whenever they appear.

Can an older adult sleep after hitting their head?
Normal sleep is allowed when a health service has said home care is appropriate and has not given different instructions. Inability to wake, abnormal drowsiness, confusion or another danger sign requires emergency help.

Should an anticoagulant be stopped?
A caregiver should not make that decision. The medicine name and last dose time should be reported promptly, and the clinician should decide whether treatment continues, pauses or changes.

Wild mushroom poisoning needs early action before symptoms appear

APPI News Editorial — Sat, 22 Aug 2026 15:05:49 GMT

The US Centers for Disease Control and Prevention (CDC) recorded 39 suspected amatoxin mushroom poisonings in Northern California from November 18, 2025, to March 17, 2026; three patients needed liver transplants and four died. The report, published May 28, 2026, described the largest mushroom-associated hepatotoxic outbreak recorded in California.

The outbreak involved suspected amatoxin-containing Amanita mushrooms, not Chlorophyllum molybdites. The two can produce different timelines: C. molybdites is linked to rapid gastrointestinal illness, while amatoxin symptoms often begin more than six hours after ingestion. Timing alone cannot identify a mushroom or make home observation safe.

Suspected ingestion needs advice before symptoms start

Australia's New South Wales Health advises immediate poison-center contact after suspected wild-mushroom ingestion and says not to wait for symptoms. Its guidance gives a broad onset range of 30 minutes to 24 hours, depending on the species, amount eaten and general health. Outside Australia, the appropriate contact is the local poison center or urgent medical service, not the Australian number.

The World Health Organization (WHO) describes poison centers as services that advise on the diagnosis and management of poisoning and maintains a directory by country; only 47 percent of WHO member states had a center as of January 1, 2023. Where no public poison center is available, the local emergency medical service or hospital is the practical route. Service names and telephone numbers differ by country.

Call local emergency services immediately if the person stops breathing, collapses, has a seizure, develops severe breathing difficulty or cannot be awakened. Australia's Healthdirect lists breathing difficulty, blue lips, confusion, marked drowsiness, unconsciousness and seizures among serious poisoning signs. Repeated vomiting that prevents fluid intake, very little urine, blood in the stool, altered alertness or severe abdominal pain also needs prompt medical assessment.

Appearance and cooking cannot establish safety

Australia's Queensland government says there is no simple, reliable test for a mushroom's safety and that even specialists can struggle because appearance changes with growth. Appearance, taste or a single photograph is therefore not a safety check. A photograph becomes useful after an exposure only as evidence for poison specialists or mycologists.

New South Wales Health says cooking, soaking, peeling or drying wild mushrooms does not remove their poison. Queensland gives the same warning. A cooked dish still counts as a suspected exposure when the mushroom was not reliably identified before use.

Chlorophyllum molybdites can cause rapid gut symptoms

Queensland's species guidance describes Chlorophyllum molybdites, or green-spored parasol, as a lawn and garden mushroom whose initially white gills turn greenish as spores mature; it says symptoms usually begin within 30 minutes to two hours. The listed effects include severe vomiting and diarrhea, abdominal pain, possible intestinal bleeding, and substantial fluid and electrolyte loss. Those traits describe known specimens but do not form a safe foraging checklist.

The US Food and Drug Administration separately lists C. molybdites among gastrointestinal irritants with rapid-onset nausea, vomiting, diarrhea and abdominal cramps, and says vomiting or diarrhea can last for several days. Its handbook identifies dehydration and electrolyte imbalance as the main dangers from this group, particularly for very young, very old or debilitated patients. An onset inside the reported range supports concern but does not confirm the species.

Four steps after a suspected ingestion

The safest response is to preserve information and let a poison service decide the next step. These steps apply whether the mushroom was raw or cooked and whether symptoms have appeared.

Stop the meal and prevent further exposure. Set aside the remaining mushrooms and food so nobody else eats them. Record the approximate ingestion time and amount.
Contact the local poison center or urgent medical service. Report the age and general health of each person exposed, how the mushrooms were prepared, when they were eaten and any symptoms. Do not wait for illness before making contact.
Do not induce vomiting or improvise a remedy. Queensland's guidance says not to induce vomiting or give anything to drink unless a poison center or medical professional directs it. Follow the case-specific instructions given by that service.
Keep a sample and photographs. New South Wales Health advises taking a photograph or retaining a mushroom sample to help identification. Keep any remaining cooked food and tell the service whether other diners developed symptoms.

Children and medically vulnerable people need direct advice

Children

New South Wales Health identifies young children as a higher-risk group because they may put mushrooms found in gardens or parks into their mouths. Any suspected bite or ingestion warrants an immediate poison-center call even when the amount is uncertain and no symptoms are present. Vomiting should not be induced.

Pregnancy

APPI News could not find one internationally adopted pregnancy-specific first-aid sequence for wild-mushroom poisoning at the time of writing. A pregnant person with a suspected ingestion should contact a poison center and the maternity or medical clinician managing the pregnancy without waiting for symptoms. Collapse, a seizure, breathing difficulty or loss of consciousness still requires local emergency services.

Older adults, chronic illness and regular medicines

The US Food and Drug Administration says dehydration and electrolyte imbalance from gastrointestinal mushroom poisoning pose particular danger to very old or debilitated patients. Older adults and people with chronic illness or regular medicines should report those details at the first call to a poison specialist or clinician. General hydration or symptom-relief advice should not replace individual directions.

Frequently asked questions

Can a wild mushroom be cleared for eating from a photograph?

No. A photograph may help a poison service or mycologist assess an exposure, but it cannot prove that a mushroom is edible before a meal. Similar-looking species and changes during growth make image matching unreliable as a safety decision.

Does cooking make a poisonous mushroom safe?

No. Official guidance from New South Wales and Queensland says cooking, soaking, peeling or drying does not make a poisonous wild mushroom safe. Heating a mushroom does not change the need for poison-center advice after a suspected ingestion.

Should vomiting be induced after a wild mushroom is eaten?

No. Induced vomiting can cause harm and should not be attempted unless a poison specialist or medical professional gives case-specific instructions. Contact the local poison center or urgent medical service instead.

Does feeling well after six hours rule out poisoning?

No. Some mushroom syndromes begin quickly, while the CDC says gastrointestinal symptoms from amatoxin poisoning often start more than six hours after ingestion. A symptom-free interval does not replace poison-center or clinical assessment.

WHO prioritizes highest-risk adults for routine COVID-19 vaccination

APPI News Editorial — Sat, 22 Aug 2026 12:00:00 GMT

The World Health Organization (WHO) published a new position paper on COVID-19 vaccines on July 31, 2026. The paper replaces earlier interim guidance and directs routine vaccination toward people at the highest risk of severe disease.

The guidance does not produce one answer for every adult. A current decision depends on the national program, time since the last dose or infection, age, health conditions, pregnancy, medicines and the products authorized where vaccination takes place.

Breathing difficulty or chest pressure needs immediate care

WHO lists difficulty breathing at rest, an inability to speak in sentences, persistent chest pain or pressure, confusion, unusual drowsiness, loss of consciousness, cold or clammy skin that turns pale or blue, and loss of speech or movement as severe COVID-19 warning signs. Anyone with these signs needs immediate medical attention through the emergency service or urgent-care facility where they are located.

The US Centers for Disease Control and Prevention says an acute allergic reaction after vaccination can involve widespread hives, throat swelling, wheezing, low blood pressure or shock. Such reactions can begin within minutes, and breathing difficulty, shock or collapse requires immediate emergency help.

WHO separates the highest-risk groups from everyone else

WHO asks countries to consider routine vaccination first for the oldest adults, older adults with major health conditions or severe obesity, residents of care homes and long-term care facilities, and moderately or severely immunocompromised people from six months of age. The agency says countries should set the age thresholds; it suggests 75 or 80 years for the oldest group and 60 years for older adults with major risk conditions.

For these highest-risk groups, WHO recommends at least one dose a year and preferably two about six months apart when the previous dose was more than six months ago. Each country still decides whether to offer one or two doses after considering local epidemiology, cost-effectiveness and delivery capacity.

Countries may also consider routine vaccination for older adults without major risk conditions, younger people with major health conditions or severe obesity, and health or care workers who have direct contact with people at high risk. Healthy adults aged 18 to 59 without those risk factors are not among the routine target groups listed in WHO's 2026 table, although a national program may use broader eligibility.

WHO found only five recent vaccine-effectiveness studies in healthy adults aged 18 to 64, four from the United States and one spanning four Western European countries. The limited and geographically concentrated evidence does not support one worldwide estimate of benefit for a healthy younger adult.

Earlier doses still count, but protection changes

Earlier vaccination remains part of a person's immune history, but it does not create a permanent exemption from later doses. WHO says protection against severe disease and death is highest in the first months after vaccination and then decreases, with immunity declining relatively quickly within six months after vaccination or infection.

WHO's vaccine composition advisers recommended a monovalent LP.8.1 antigen in May 2026 to improve immune responses to circulating variants, while accepting other antigens or approaches that show broad responses or efficacy. The same statement says people at high risk should not delay vaccination while waiting for a newer formulation; a composition recommendation does not establish that a product is authorized or stocked in a particular country.

WHO also says gaps and delays in countries' reporting of cases, hospitalizations and deaths limit comparisons of COVID-19 trends. A rise reported in one country therefore cannot establish that infections are increasing worldwide.

Pregnancy has its own recommendation

WHO recommends one COVID-19 vaccine dose during each pregnancy, at any stage and ideally during the second trimester. The aim is to protect the pregnant person from severe disease, reduce adverse pregnancy outcomes and transfer maternal antibodies to the infant.

WHO's evidence review found no safety concern from COVID-19 vaccination during pregnancy across observational studies, pregnancy registries and vaccine surveillance systems in multiple countries. It also reported no safety concern for breastfeeding people or their infants. Breastfeeding people at higher risk follow the recommendation for their risk group, while the locally authorized product and national schedule still apply.

Immune suppression, chronic illness and medicines change the plan

WHO includes older adults with major cardiovascular, respiratory, kidney, liver or neurological conditions, diabetes or severe obesity among the highest-risk groups. Younger adults with these conditions are an additional group that countries may include, so the same age does not produce the same recommendation for every person.

WHO includes people receiving immune-suppressing treatment in a higher-risk group, so the condition, treatment plan and vaccine record belong in the assessment. WHO says anaphylaxis after a previous dose or to a vaccine component rules out further doses of the same vaccine type, while a history of vaccine-associated myocarditis or pericarditis requires an individual risk-benefit assessment.

Healthy children do not use the adult schedule

WHO does not routinely recommend repeat COVID-19 vaccination for healthy children or adolescents. Countries may consider primary vaccination for previously unvaccinated healthy children aged six to 23 months only when they document a substantial disease burden in that age group.

Moderately or severely immunocompromised children from six months of age fall into a higher-risk group for which WHO asks countries to consider routine vaccination. Pediatric product authorization and schedules vary by country, so an adult interval cannot be applied to a child.

Other vaccines can be given at the same visit

WHO says a COVID-19 vaccine may be given with, before or after other vaccines, including vaccines recommended during pregnancy. Vaccines given during the same visit should use separate injection sites, preferably on different limbs.

Randomized trials, systematic reviews and post-authorization monitoring found no clinically meaningful reduction in immune responses or vaccine effectiveness when COVID-19 and other routine vaccines were given together. Some studies found small, temporary increases in injection-site pain, fatigue or muscle aches, while national product instructions still govern each combination.

Common reactions usually settle within days

WHO lists fever, tiredness, headache, muscle aches, chills, diarrhea, and pain or redness at the injection site among expected reactions. Most resolve within a few days, while serious or long-lasting reactions are possible but extremely rare.

A vaccination check should record the date and product used for previous doses, the date of a confirmed infection, current health conditions and medicines, pregnancy, and any severe reaction after a previous dose. The vaccinating service can then apply the current national schedule and the instructions for products authorized in that country.

APPI News did not verify authorization, supply, price or public coverage market by market at the time of writing. It also could not verify a comparable worldwide percentage for the source article's Taiwan-specific "more than 90 percent" figure, so that number is omitted. A public campaign deadline in one country cannot be used as an eligibility date elsewhere, so those details must come from the applicable national immunization program or an authorized vaccination provider.

Frequently asked questions

Does someone who received several earlier doses still need another?
Possibly. WHO's 2026 schedule focuses on current risk group and time since the last dose, while each country sets eligibility and timing.

Is a six-month interval a universal rule?
No. WHO prefers an interval of about six months for routine vaccination and recommends at least one dose a year, preferably two, for the highest-risk groups, but countries decide the number and timing.

Can COVID-19 and influenza vaccines be given together?
WHO accepts vaccination during the same visit with separate injection sites. The products must still be authorized for the person and used according to the schedule in that country.

Does WHO routinely target a healthy adult under 60?
Not in its July 2026 table if the adult has no major risk condition. A country's program may set broader eligibility based on local epidemiology and delivery priorities.

Should a high-risk person wait for a newer formulation?
WHO says no. Its May 2026 composition statement says vaccination should not be delayed while waiting for an updated antigen, although product authorization and availability vary by country.

Ajinomoto reports strong ABF sales as supply-cut claim remains unverified

APPI News Editorial — Fri, 21 Aug 2026 18:05:35 GMT

Japan's Ajinomoto Co. said on August 6, 2026, that sales of Ajinomoto Build-up Film (ABF) remained strong for high-performance boards used in AI systems, servers and networks. A June 30 briefing by its electronic-materials subsidiary, Ajinomoto Fine-Techno Co., said the company had no concerns about its current supply chain overall.

Those disclosures predate an August 20 claim cited by the source report that some customers in China could receive roughly 30 percent less ABF. APPI News could not find an Ajinomoto release or investor filing that confirmed the customer group, percentage or timing as of August 23, 2026.

The distinction matters because ABF is embedded in the manufacturing process for high-end processor substrates. A real shortage would first affect qualified film grades and substrate output, while an unverified allocation claim cannot be converted directly into chip delays or new orders for another supplier.

ABF is the film, not the substrate

An ABF substrate is a multilayer package substrate built with ABF between its copper routing layers. Ajinomoto Fine-Techno describes ABF as an insulation film that is laminated onto a substrate before wiring is formed through later processes.

The substrate sits between a semiconductor die and the main circuit board, spreading a dense set of chip connections across a larger area. Taiwan-based Kinsus Interconnect Technology says flip-chip ball-grid-array substrates serve high-pin-count microprocessors, graphics processors, application-specific integrated circuits and field-programmable gate arrays.

Ajinomoto says ABF combines epoxy resin, a hardener and inorganic microparticle filler in a thermosetting film whose surface supports laser processing and direct copper plating, and it dates the film's first adoption by a major semiconductor maker to 1999. Those are the supplier's own material and product-history claims.

Market share and qualification make substitution slow

Ajinomoto Fine-Techno puts ABF's share near 100 percent in the insulation-film market for major personal computers worldwide. Its June 30 briefing separately claimed more than 95 percent of what it called ABF's market, but neither disclosure identifies the competing suppliers or publishes a measurement method.

Ajinomoto says ABF is custom-manufactured for each customer and shows development moving from laboratory work through customer evaluation before mass production. That disclosure supports an inference, rather than a published lead-time estimate, that changing film requires a customer to requalify the material and process.

A package-substrate maker must control lamination, laser-drilled vias, copper plating, fine routing, warpage and reliability for the selected film grade. A material with similar specifications is therefore not an interchangeable production source until it passes those manufacturing and customer checks.

The supply chain has more than one capacity point

Ajinomoto Fine-Techno says it makes ABF varnish internally at plants in Kawasaki and Gunma, Japan, while contractors coat and cut the film, store it frozen and ship it. Capacity can therefore tighten in raw materials, varnish production, film conversion, cold storage, logistics or substrate fabrication.

The same briefing says a new varnish plant at Gunma entered full operation in 2025 and that the company plans staged production increases through 2030. Ajinomoto also plans to begin building a third production base in Gifu, Japan, in 2028 and start operations there in 2032.

Those projects show that Ajinomoto expects demand to keep growing, but they do not establish a current shortage. The company does not publish monthly output for each production step, and the planned Gifu site cannot relieve a near-term constraint.

Taiwan's substrate makers sit downstream of the film

Taiwan-based Kinsus and Nan Ya Printed Circuit Board Corp. manufacture package substrates, placing them after the film supplier in this chain. Kinsus lists fine-pitch patterns and high layer counts among the features of its flip-chip ball-grid-array substrates, but the page does not describe Kinsus as an ABF film producer.

Nan Ya PCB's published roadmap lists a 24-layer ABF substrate target for 2026 and more than 24 layers for the first half of 2027. A development roadmap does not show how much qualified capacity is available for a particular customer or whether a product has entered volume production.

Neither company's official page establishes that it received orders because of a cut by Ajinomoto. APPI News also could not verify the source report's claims about validation and small-volume shipments by a Taiwan-based alternative-film supplier in an English-language primary disclosure, so no replacement-film company is identified here.

A real shortfall would move through several gates

An actual film shortfall would reach substrate makers before chip assembly and server production. Its effect would depend on film inventories, the grades allocated to each customer, the amount used per substrate, manufacturing yield and the qualified substrate capacity available for each chip.

Ajinomoto says larger, more highly layered AI package substrates consume more ABF per unit than earlier computer packages. That raises material exposure, but a percentage change in film allocation still cannot be mapped one-for-one to processor shipments because package designs and inventories differ.

The evidence to watch is a named company disclosure that identifies the affected grade, customer group and period, followed by substrate makers reporting changes in qualified output or delivery schedules. None of those details was available in Ajinomoto's public English materials at the time of writing.

Frequently asked questions

Are ABF film and an ABF substrate the same product?
No. ABF is the insulating film placed between copper routing layers, while an ABF substrate is the completed multilayer structure that connects a chip package to a circuit board.

Has Ajinomoto confirmed a 30 percent supply cut in China?
APPI News could not find an Ajinomoto release or investor filing confirming that claim as of August 23, 2026. The company's June 30 supply statement and August 6 sales update came before the reported August 20 claim, so they do not disprove a later customer-specific change.

Why can a substrate maker not switch film immediately?
Ajinomoto says it develops customer-specific ABF and includes customer evaluation before mass production. A replacement has to work with lamination, laser drilling, plating, fine routing and package-reliability requirements before it can enter qualified production.

Do Taiwan's substrate makers automatically benefit from an ABF shortage?
No. Their ability to make large, highly layered substrates does not prove that they have spare qualified capacity or access to an interchangeable film, and no official disclosure confirms new orders tied to the reported cut.

Apple Pay sends device tokens instead of card numbers to merchants

APPI News Editorial — Fri, 21 Aug 2026 17:26:28 GMT

Apple says Apple Pay sends a device-specific account number and a transaction-specific security code instead of the original card number when an approved card is used in a store, app or website. The replacement credential narrows the card data exposed at checkout, but it does not remove the issuer, payment network or merchant from the transaction.

EMVCo defines payment tokenization as replacing the primary account number (PAN) with a unique alternative value that can be restricted to a merchant, device or payment scenario. The token still travels through the established route from the point of purchase to the acquirer, payment network and card issuer for authorization.

An issuer creates a device-specific payment credential after approving a card for Apple Pay. (Illustrative image)

Card setup creates a credential for one device

Adding a card starts a provisioning exchange rather than a simple copy of the number into Wallet. The issuer or its authorized service provider uses card, account and device information to decide whether to approve enrollment, and it may require a separate verification step.

Apple's platform-security documentation says the issuer creates a unique Device Account Number, sends it to Apple in encrypted form and stores it in the device's Secure Element; the full card number is not stored on the device or on Apple Pay servers. Apple says it cannot access the encrypted Device Account Number, which is not backed up to iCloud.

The credential is specific to the enrolled device. Apple says a card added to Apple Watch receives its own Device Account Number, while another Apple Pay-capable device provisions the card through a new exchange rather than copying credentials from the first device.

Store payments pair NFC with a one-time cryptogram

At a contactless terminal, Near Field Communication (NFC) carries payment data between the device and the point-of-sale system. Face ID, Touch ID or a passcode normally confirms the user's intent before the Secure Element releases the Device Account Number and dynamic security code, although eligible payment and transit cards can use Express Mode without that check.

Apple describes the dynamic code as a one-time payment cryptogram calculated from a transaction counter and a key provisioned to the payment applet. NFC payments can also incorporate a number supplied by the terminal, and the payment network or issuer checks the resulting code during authorization.

NFC carries the device credential and a transaction-specific security code to the terminal. (Illustrative image)

Apps and websites use a separate encrypted route

App and web checkout does not use the NFC path, but the card-number boundary is similar. Apple says it receives the encrypted transaction, reencrypts the payment information with a developer-specific key and sends the Device Account Number and transaction-specific code without sending the original card number to the app or website.

A website accepting Apple Pay must use an Apple merchant identifier, a payment-processing certificate, registered merchant domains and a merchant identity certificate. The payment-processing certificate protects transaction data, while the merchant identity certificate authenticates communication with Apple Pay servers.

The protected payment credential is not the whole order. Apple's privacy notice says a merchant may receive information it requests after authorization, including a device- or merchant-specific account number, shipping address or email address, even though the original credit, debit or prepaid card number is not provided.

Payment credentials and order information cross different data boundaries during online checkout. (Illustrative image)

Tokenization limits exposure rather than eliminating payment risk

A restricted token is less useful after a breach than an unrestricted PAN because its permitted device, merchant or transaction context can be narrower. EMVCo describes that reduction in exposure as the security benefit; it does not remove the need for the issuer to authorize a payment or for each participant to protect the data it holds.

For a merchant, the distinction creates two operational boundaries. The payment credential passes through the payment processor and network, while names, email addresses, shipping details and order records may enter the merchant's own systems and remain subject to its access, retention and privacy controls.

Availability has no single worldwide card list

Apple's setup guidance, updated July 6, 2026, requires a compatible device, an Apple Account, a supported card from a participating issuer and availability in the user's country or region. Apple directs cardholders to the issuer for card-level compatibility and says some issuers may request more information or require their app before approval.

APPI News reviewed Apple's global setup, privacy and platform-security documents but did not verify eligibility across every issuer and market. This report therefore does not reproduce a bank list or imply that every card from a participating issuer can be added.

Availability depends on the country or region, issuer, card, device and payment setting. (Illustrative image)

Lost-device controls act on the enrolled payment credentials

Apple says Lost Mode can suspend Apple Pay when Find My was enabled, while the Apple Account page or a remote erase can remove the ability to pay with cards on the missing device. A bank or issuer can also suspend credit, debit and prepaid cards from Apple Pay even when that device is offline.

Suspending the device credential is not always the same as canceling the physical card account. Because each enrolled device receives its own credential, the status of other devices using the same card must be considered separately, and Express Mode remains an exception to the usual biometric or passcode step until the relevant credential is suspended.

AI hip-pain triage needs emergency rules and human review

APPI News Editorial — Fri, 21 Aug 2026 17:26:28 GMT

The International Medical Device Regulators Forum (IMDRF) finalized 10 good machine learning practice principles for medical devices that use artificial intelligence (AI) on January 27, 2025. The principles cover intended use, representative data, human-AI performance and monitoring throughout the product lifecycle.

For a service that sorts reports of hip pain, those controls point to a limited role: collect structured information, identify missing answers and propose an urgency class for review. The software cannot perform a physical examination or establish why a joint hurts.

APPI News could not find a publicly available independent clinical evaluation of an AI system specifically used to triage hip pain in Taiwan at the time of writing. The available evidence supports only a workflow analysis, not a claim that a validated product is available in Taiwan or elsewhere.

Emergency signs must bypass routine automation

England's National Health Service (NHS) says adults need immediate emergency care for severe hip pain after a fall or injury, inability to walk or bear weight, or tingling or loss of feeling after injury. It advises urgent medical assessment for sudden severe pain without injury, a hot or swollen hip, skin color change, or hip pain with fever or feeling unwell. People with these signs should use local emergency or urgent-care services rather than wait for a chatbot or routine booking process.

A patient-facing system should evaluate these warning signs before a model ranks departments or appointment slots. If a sign is present, the software should stop routine routing, state the trigger and hand the case to a human-controlled urgent pathway; model confidence must not suppress the escalation.

The NHS page covers adults and directs readers to separate guidance for children. Children, pregnant patients, people with chronic conditions and people taking regular medicines should speak to a clinician rather than rely on a general adult triage pathway, because APPI News did not find one rule set validated across these groups and national care systems.

Published evidence supports a narrower task

A 2020 feasibility study in JMIR Medical Informatics analyzed 576 complete cases from adults with knee or hip pain who had already been referred by general practitioners within one health board in Wales. The researchers used topics extracted from referral letters to predict subsequent treatment outcomes and evaluated the classifiers with 10-fold cross-validation.

The study tested referral-letter routing after a clinician had entered the care pathway. It did not evaluate a consumer symptom chatbot, acute warning-sign detection, a hip-only cohort or external performance in another health system, so its feasibility result cannot establish safe autonomous triage.

A bounded output routes cases instead of diagnosing them

Structured intake can record onset, injury, ability to walk or bear weight, pain location, fever or feeling unwell, swelling, heat or skin change and the effect on sleep or daily activity. The output can then show urgency, the destination class available in that health system, unanswered safety fields and the reason for the route.

Department names and referral rights differ by country and provider. A global service should use categories such as emergency, urgent clinical assessment or routine assessment until a local operator maps them to its own departments; it should not present orthopedics or rehabilitation medicine as a universal direct-access route.

Under US policy, the US Food and Drug Administration (FDA) says software meant to support time-critical decision-making does not meet the country's non-device clinical decision support criteria because clinicians may lack time to independently review the basis of a recommendation. The agency gives automatic retrieval of relevant patient information when a clinician places an order as an example that may still qualify in a time-critical setting. Failing one exclusion criterion does not automatically make every such product a regulated device, and the US classification does not determine status elsewhere.

FHIR carries structured data, not accountability

Health Level Seven International (HL7) defines Fast Healthcare Interoperability Resources (FHIR) Release 5 as a standard for electronic exchange of health information, with Resource as its basic building block. Its modules cover clinical content, observations and reports, referrals, workflow and clinical decision support. A hip-pain workflow could use those structures to pass intake answers and routing records between applications without treating a whole note as one text block.

HL7 also says FHIR is not a security protocol and that a production deployment needs systems for authentication, authorization and audit. FHIR supplies structures such as security labels, Provenance and AuditEvent, but implementers remain responsible for deciding who may read or change a triage result.

A valid Resource does not prove that the source answer is complete, correctly coded or current. Hospitals still need local profiles, terminology rules, validation, access limits and a way to correct records before a model uses them.

Four controls come before clinical deployment

First, the intended use must be defined. The product record needs to name the patient population, users, care setting, inputs, output, decision influenced, exclusions, response time and fallback when data are missing or the service fails.

Second, testing must cover the actual workflow. The IMDRF principles call for representative training, test and monitoring data, independent training and test sets, clinically relevant conditions and evaluation of the human-AI team rather than the device alone. Evidence from one adult referral cohort cannot be extended without testing to children, acute injuries, postoperative patients or different language and care settings.

Third, the result must remain inspectable and overridable. Joint principles issued by the US Food and Drug Administration, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency (MHRA) say transparency should cover intended users and populations, inputs and outputs, effects on clinical judgment, training and testing data, known gaps, performance monitoring and change management. For a triage service, that supports displaying source answers, missing fields, the reason for urgency and the model version to a reviewer; this four-field screen is an APPI News synthesis, not a regulator-mandated template.

Fourth, the service needs an audit trail and monitoring after release. Each record should preserve the system version, input state, recommendation, human disposition and reported failure, while site-level review tracks shifts in missing data, overrides and errors across patient groups. These are controls around a clinical service, not proof of product accuracy.

What remains unverified

APPI News could not verify a publicly available independent evaluation or deployment record for AI hip-pain triage in Taiwan. That search gap does not prove that no local pilot or product exists; the accurate finding is “not verified,” not “absent.”

The 2020 Wales study addressed treatment routing from clinician-written referral letters, while the international documents describe development practices and do not approve a particular system. A product's evidence, intended use, availability and regulatory status must be checked in each country; none of the cited documents establishes that a standalone hip-pain chatbot is safe for autonomous use.

Australian police charge dual citizen in alleged Ukraine intelligence case

APPI News Editorial — Fri, 21 Aug 2026 15:55:42 GMT

The Australian Federal Police (AFP) charged Russian-Australian dual citizen Vladimir Teslov, 27, on August 21, 2026, with one count of attempting to engage in intentional foreign interference. Brisbane Magistrates Court in Australia adjourned the matter until October 2. Teslov has not been convicted, and the 20-year maximum penalty is not a sentence already imposed.

Police allege Teslov joined Ukraine’s armed forces to collect information about military personnel, units and locations, then provided or attempted to provide it to people he believed were acting for Russian intelligence services. The AFP said the alleged conduct created a risk to Ukrainian military personnel but that investigators had found no indication of a threat to Australia.

Police allege a timeline across three countries

The AFP alleges Teslov traveled to Russia in October 2024 for military-style training and returned to Australia in December. Investigators say he remained in contact with people he believed were connected to Russian intelligence services. They allege he traveled to Ukraine in May 2025, joined its armed forces and gained access to operational information.

Australia’s Counter Foreign Interference Task Force opened Operation Woodcreeper in June 2026 after an extensive investigation by the Australian Security Intelligence Organisation (ASIO). Officers executed five search warrants in and around Brisbane and seized electronic devices and other items for forensic examination. AFP Commissioner Krissy Barrett said investigators had not established whether Teslov worked with associates in Australia and that further arrests were possible.

What the foreign-interference charge requires

Section 92.2(1) of Australia’s Criminal Code applies when conduct is carried out on behalf of or in collaboration with a foreign principal, or is directed, funded or supervised by one. The person must intend one of four listed outcomes, including support for the principal’s intelligence activities, and some part of the conduct must be covert, involve deception, threaten serious harm or involve a demand with menaces. Supporting a foreign principal’s intelligence activities is a separate statutory route from prejudicing Australia’s national security. Section 92.6 gives offenses in this division extended geographical jurisdiction, while the AFP alleges that parts of Teslov’s conduct occurred in Australia and overseas.

Section 11.1 says an attempt must go beyond conduct that is merely preparatory and is punishable as if the attempted offense had been completed. Section 92.2(1) sets a 20-year maximum, which therefore also applies to the attempted offense alleged in this case. A statutory maximum does not predict the sentence a court would impose after a conviction.

The police release identifies the statutory provisions but does not present the prosecution evidence for each element. APPI News could not find a publicly available charge sheet or prosecution filing with those details as of August 23, 2026. The case is legally distinct from David McBride’s convictions for taking and unlawfully communicating Australian military information, which did not involve a foreign-interference charge.

Evidence and possible associates remain unresolved

The AFP says devices and other seized items still require extensive forensic examination. Its public account does not identify the military material at issue or distinguish information allegedly transmitted from information covered only by an alleged attempt. Police have also not said whether any other person will be charged.

The Russian Embassy in Canberra told ABC News that it learned of the case from media reports and had received no official request or notification from Australian authorities or Teslov. APPI News could not independently verify communications between the embassy, Australian authorities and the accused.