VA Providence began rolling out an ambient AI scribe in primary care on March 24, 2026, with participation optional for veterans. The software turns clinical conversations into draft notes, leaving accuracy, consent and liability subject to rules that vary by country.
The 2019 Apple Heart Study sent irregular-pulse notifications to 0.52 percent of 419,297 participants, while later ECG patches found atrial fibrillation in 34 percent of 450 analyzable recipients. Its 84 percent positive predictive value measured a different event: another alert that occurred during simultaneous ECG recording.
National Taiwan University Hospital said in 2026 that it had built a federated-learning platform with partner hospitals to validate medical AI without moving their data off site. The approach limits centralized data collection, but it does not eliminate privacy, performance or accountability risks.
The US Food and Drug Administration cleared AEYE-DS for use with the handheld Optomed Aurora retinal camera on April 23, 2024. The system automatically returns a screening result, but its authorization and performance data have defined limits.
UC San Diego researchers reported on July 8, 2026, that a surgeon remotely controlled a humanoid robot during two gallbladder operations on pigs. The study showed technical feasibility, but human bedside assistance, repeated recalibration and sterility problems remain barriers to clinical use.
Taiwan added 17 smart assistive products to a long-term-care rental benefit on July 1, 2026. The published list identifies products and suppliers but does not provide performance metrics or explain how each system makes decisions.
OpenAI began rolling out Health in ChatGPT to US adults on July 23, 2026, with connections to Apple Health and supported medical records. Taiwan's My Health Bank SDK instead sends selected records to third-party apps after a government review, and new SDK applications are currently paused.
Researchers in South Korea and the United States reported two manufacturing advances for stretchable electronics in 2026. One kept liquid-metal conductors working under repeated strain, while the other put a dense computing array on a stretchable patch, but neither study tested a finished implant in people.
A systematic review published in January 2026 found that generative AI increased the volume, speed and perceived credibility of health misinformation in the available studies. Clinicians can use the same tools to draft public information, but the evidence and professional guidance leave accuracy, privacy and disclosure with the person who publishes it.
A July 30, 2026 Coldcard advisory showed how a random-number integration error could remain hidden while security-sensitive outputs still looked valid. The same failure pattern matters in health information systems that generate pseudonyms and SMART on FHIR authorization secrets.
A September 2025 paper argued that language-model training and widely used benchmarks reward guessing over acknowledging uncertainty. A 2026 health-chatbot audit found 49.6 percent of answers problematic, showing how fluent errors can reach high-stakes uses.
Taiwan-listed Bionime secured a local medical-device license for its iFree2 continuous glucose monitor in April 2026, but the public notice did not identify an AI forecasting feature. Abbott's Libre Assist offers such predictions in the United States, while Abbott's Taiwan product site lists an older sensor and does not advertise the feature.
Taiwan's health insurance agency said in August 2024 that it aimed to finish a cost-effectiveness review of AI diagnostic tools by the end of that year. APPI News could not find a published payment decision for the proposed CT brain-scan application as of August 3, 2026.
A 2025 laboratory study found that six wrist-worn trackers identified more than 90 percent of sleep intervals but correctly identified only 29.39 percent to 52.15 percent of wake intervals. Their movement and pulse sensors can show broad patterns, but their sleep-stage estimates do not match direct physiological measurements consistently.
The US Food and Drug Administration cleared UpDoc on December 23, 2025, as prescription software for adults with type 2 diabetes. Its chat interface collects patient data, while insulin instructions follow clinician-set treatment parameters.
Oregon enacted an AI companion law in 2026 that requires crisis protocols and allows harmed users to sue. Illinois, Nebraska, Tennessee and Maine have taken different approaches, leaving the rules dependent on a product's claims, design and users.
Taiwan's health ministry says three hospital groups completed a cross-system FHIR demonstration by the end of 2025. The standard can make records easier for software and AI tools to query, but it does not settle data quality, access control or accountability.
The World Health Organization and the International Telecommunication Union mapped the use of AI in traditional medicine in a 2025 technical brief. Their findings, combined with international medical-software standards, show why a product demonstration cannot establish clinical credibility.
A 2024 systematic review found that explainable AI increased clinicians' trust in five of 10 empirical studies, while two reported mixed effects. A COVID-19 drug-repurposing project shows how an inspectable workflow can support expert review, but explanations do not establish clinical validity.
International guidance gives health organizations a practical set of controls to test before deploying generative AI. The work starts with data handling and audit design, then extends to source traceability, human authority and continuous monitoring.
Taiwan Health Nexus began operations on January 9, 2026, with plans to connect health records held by separate institutions for AI and research. Its public claims address interoperability, but evidence on consent, oversight and commercial reuse remains limited.
Taiwan launched TwHealth Nexus on January 9, 2026, with plans to connect health records held by hospitals and other care providers. The project goes beyond Taiwan's patient-facing My Health Bank service by proposing secondary data use for research and AI, raising separate questions about consent, withdrawal and oversight.
GPT-5 Pro generated a mechanism for unexplained T-cell data and predicted the direction of an unpublished CAR-T experiment, according to a November 2025 case study. The result shows how a model can help experts form testable hypotheses, but it does not establish autonomous discovery or prove the proposed mechanism.
OpenAI and Molecule.one reported on June 17 that a GPT-5.4 workflow helped identify conditions that improved a difficult Chan-Lam coupling across 10,080 microscale reactions. An automated laboratory supplied the testing capacity, while chemists selected proposals, corrected plans and repeated key experiments by hand.
