Watch

Signals worth watching in health AI

A running feed of the regulatory, legal, and safety signals shaping health-AI governance, each with its source and why it matters.

Alert

Former Research Director Sues Mayo Clinic, Claims the Health Network Deliberately Conceals Failure Rates in Deployed AI Tools

Mayo Clinic; U.S. District Court for the District of Minnesota

Traci Tamiko Eto, a former Mayo Clinic AI leader, is alleging that the health system concealed error rates and integrated surgical devices without review board approval. According to the lawsuit, after Eto escalated concerns, her employment was terminated.

Alert

Xsolis Reports 1.3 Million Patient Records Exposed After Phishing Attack

Xsolis

Last month, Xsolis, used by over 600 hospitals, revealed that threat actors gained access to 1.3 million patient SSNs. Its response reveals an industry-wide security doctrine, crippling for providers.

Clinical AI evaluation concept for the OpenEvidence and Nature study dispute

Risk

OpenEvidence Disputes a Nature Study Reporting Frontier LLMs Outperformed Clinical AI Tools

Nature (study); OpenEvidence (LinkedIn response); NYU Langone Health

A Nature study found frontier LLMs matched or beat clinical AI tools across three evaluations. OpenEvidence's accusatory response raises its own questions about transparency and evaluation.

Vials and medical supplies illustrating hospital drug-diversion monitoring

Risk

Sentri7 Failed to Detect Fentanyl Diversions at a Tennessee Hospital

Wolters Kluwer (Sentri7); Tennessee Board of Nursing; CDC

Wolters Kluwer's Sentri7, used by 700+ hospitals, missed five fentanyl diversions at Erlanger Baroness — caught only by chance human observation. The case for continuous post-deployment monitoring.

Genomic data visualization for AI-assisted rare-disease reanalysis

Alert

Beyond the Diagnostic Odyssey: AI-Assisted Genomic Reanalysis in Rare Disease

Harvard Medical School, Boston Children's Hospital & OpenAI (joint study)

A Harvard/Boston Children's/OpenAI study found 18 new diagnoses across 376 unsolved rare-disease cases using AI-assisted genomic reanalysis — the model as a hypothesis engine, clinicians as adjudicators.

Compliance

The EU AI Act's high-risk healthcare obligations hit a 2026 milestone

Tandem Health

The EU AI Act's high-risk obligations for healthcare AI arrive around August 2026, and US systems can inherit them through EU-facing vendors.

Abstract AI visualization representing clinical prediction models

Risk

Deployed sepsis-prediction models degrade after go-live

Frontiers in Medicine

A peer-reviewed review finds deployed sepsis-prediction models, including Epic's, degrade after go-live, making post-deployment revalidation essential.

Abstract AI concept representing unmonitored AI use

Alert

Nearly one in five clinical staff report using unauthorized AI tools

Healthcare Dive

A Wolters Kluwer survey finds nearly one in five clinical staff have used unauthorized AI tools, exposing the gap between written policy and the floor.

Abstract AI governance and regulation concept

Compliance

Texas now requires disclosing AI use in diagnosis and treatment

Spencer Fane

Texas now requires providers to tell patients when AI is used in their care, with civil penalties up to $200,000 per violation.

HHS HTI-5 proposed rule on clinical AI transparency

Regulatory

A proposed federal rule would remove the model card requirement for clinical AI

Nelson Mullins

A proposed HHS rule (HTI-5) would drop the model-card transparency mandate for clinical AI, shifting the due-diligence burden onto health systems.

Lady Justice with scales, representing the AI claim-denial court ruling

Legal

A federal court let the AI claim-denial case against UnitedHealth move forward

DLA Piper

A federal court let breach claims proceed over UnitedHealth's nH Predict AI denials, signaling that 'the algorithm decided' is no liability shield.

Abstract AI-enabled medical device technology concept

Regulatory

FDA wants AI devices watched for their whole life, not just at clearance

Greenberg Traurig

FDA draft guidance asks manufacturers to monitor AI-enabled devices across their whole lifecycle, not just at clearance.