AI policy in medicine, explained in plain language
MedAI Update tracks the rules and decisions that shape how AI is used in healthcare: FDA and HHS documents, GAO reports, state bills and the reporting that explains them. Federal items carry the agency's own summary, state bills carry the LegiScan summary, and news items name the outlet. Every item links to its source.
In plain words: A Government Accountability Office (GAO) report says covert testing with prompts to Claude, Grok, ChatGPT and Gemini produced viable attack plans, including for an improvised explosive device and the electrical grid.
Excerpt from the agency summary: In spring 2026, GAO’s prompts to four of the most used generative artificial intelligence (AI) platforms—Claude, Grok, ChatGPT, and Gemini—produced feasible information that could assist a malicious actor with planning and executing attacks. Specifically, in response to prompts, two platforms produced information for constructing and detonating an improvised explosive device (IED) designed to cause mass casualties. In addition, after prompting, two other platforms produced information for targeting the electrical grid with a drone and firearms, respectively, to cause economic harm.
In plain words: The Food and Drug Administration (FDA) classifies the digital cervical cytology slide imaging system with artificial intelligence algorithm into class II with special controls, which FDA says provides reasonable assurance of safety and effectiveness.
Excerpt from the agency summary: The Food and Drug Administration (FDA) is classifying the digital cervical cytology slide imaging system with artificial intelligence algorithm into class II (special controls). The special controls that apply to the device type are identified in this order and will be part of the codified language for classification of the digital cervical cytology slide imaging system with artificial intelligence algorithm. We are taking this action because we have determined that classifying the device into class II will provide a reasonable assurance of the safety and effectiveness of the device.
In plain words: National Institutes of Health (NIH) Division of Program Coordination, Planning, and Strategic Initiatives seeks comment on creating Office of the NIH Chief Artificial Intelligence Officer and moving the Office of Dietary Supplements under Nutrition Research.
Excerpt from the agency summary: The Division of Program Coordination, Planning, and Strategic Initiatives (DPCPSI) in the Office of the Director (OD), National Institutes of Health (NIH) is seeking public comment regarding its proposal to establish the Office of the NIH Chief Artificial Intelligence Officer (OCAIO) within DPCPSI and move the Office of Dietary Supplements (ODS) under the Office of Nutrition Research (ONR) within DPCPSI in OD, NIH.
In plain words: Food and Drug Administration (FDA) issues rule putting cardiovascular machine learning-based notification software in class II with special controls, which FDA says provides reasonable assurance of safety and effectiveness.
Excerpt from the agency summary: The Food and Drug Administration (FDA) is classifying the cardiovascular machine learning-based notification software into class II (special controls). The special controls that apply to the device type are identified in this order and will be part of the codified language for classification of the cardiovascular machine learning-based notification software. We are taking this action because we have determined that classifying the device into class II will provide a reasonable assurance of the safety and effectiveness of the device. We believe this action will also enhance patients' access to beneficial innovative devices, in part by reducing regulatory burdens.
In plain words: Food and Drug Administration (FDA) issues rule putting radiological machine learning-based quantitative imaging software with predetermined change control plan in class II with special controls, which FDA says provides reasonable assurance of safety and effectiveness.
Excerpt from the agency summary: The Food and Drug Administration (FDA) is classifying the radiological machine learning-based quantitative imaging software with predetermined change control plan into class II (special controls). The special controls that apply to the device type are identified in this order and will be part of the codified language for classification of the radiological machine learning-based quantitative imaging software with predetermined change control plan. We are taking this action because we have determined that classifying the device into class II will provide a reasonable assurance of safety and effectiveness of the device.
In plain words: The Food and Drug Administration extends the comment period for request for information on proposed pilot program to assess how artificial intelligence-enabled technologies can improve efficiency, speed, and quality of decision-making in early-phase clinical trials.
Excerpt from the agency summary: The Food and Drug Administration (FDA or the Agency) is extending the comment period for the notice entitled "AI-Enabled Optimization of Early-Phase Clinical Trials Pilot Program; Request for Information" that appeared in the Federal Register of April 29, 2026. In the notice, FDA requested comments to solicit input on a proposed pilot program to assess how artificial intelligence (AI)-enabled technologies can improve efficiency, speed, and quality of decision-making in early phase clinical trials. The Agency is taking this action in response to a request for an extension to allow interested persons additional time to submit comments.
In plain words: A Virginia bill, SB269, would permit mental health service providers to use AI for therapy or counseling if they retain full responsibility while prohibiting AI-only services and setting a civil penalty up to $10,000.
Bill title and LegiScan summary: Use of artificial intelligence system by mental health service providers; civil penalty. Permits the use of an artificial intelligence system by mental health service providers to assist in providing therapy or counseling services if such mental health service provider maintains full responsibility for all interactions, outputs, and data use associated with the system. The bill prohibits the use of an artificial intelligence system to provide therapy or counseling services without a mental health service provider.
In plain words: A Virginia state bill addresses health carriers’ use of artificial intelligence and calls for related disclosures, according to its title.
Bill title and LegiScan summary: Health carriers; use of artificial intelligence, disclosures.
In plain words: A Texas bill, HB273, would relate to the use of artificial intelligence by health care providers, and its title gives no further detail.
Bill title and LegiScan summary: Relating to the use of artificial intelligence by health care providers.
In plain words: A Texas bill, HB588, would create a health care facility grant program supporting the use of artificial intelligence technology in scanning medical images for cancer detection.
Bill title and LegiScan summary: Relating to a health care facility grant program supporting the use of artificial intelligence technology in scanning medical images for cancer detection.
In plain words: A Pennsylvania bill would provide for artificial intelligence in facilities, insurer use and MA or CHIP managed care plans, imposing duties on the Department of Health, Insurance Department and Department of Human Services and penalties.
Bill title and LegiScan summary: An Act amending Titles 35 (Health and Safety), 40 (Insurance) and 67 (Public Welfare) of the Pennsylvania Consolidated Statutes, providing for artificial intelligence in facilities, for artificial intelligence use by insurers and for artificial intelligence use by MA or CHIP managed care plans; imposing duties on the Department of Health, the Insurance Department and the Department of Human Services; and imposing penalties.
In plain words: A District of Columbia bill, B26-0853, as introduced, would prohibit health insurers from automatically downcoding a claim using artificial intelligence based solely on the reported diagnosis code and require clinician notice and appeals.
Bill title and LegiScan summary: As introduced, Bill 26-853 would prohibit health insurers from automatically downcoding a claim using artificial intelligence based solely on the reported diagnosis code. An insurer changing a service code submitted by a clinician to a lower-level code, can result in lower reimbursement. It requires they notify the clinician to clearly indicate the claim was downcoded. It provides clinicians with a process for appealing downcoded claims. It makes violations of this act enforceable by the Department of Health and Human Services.
In plain words: A California bill would add Section 1714.48 to the Civil Code and Article 2.7 to the Labor Code, relating to health care services and artificial intelligence.
Bill title and LegiScan summary: An act to add Section 1714.48 to the Civil Code, and to add Article 2.7 (commencing with Section 2820) to Chapter 2 of Division 3 of the Labor Code, relating to health care services.
In plain words: A California bill would add Chapter 13.6 to the Business and Professions Code, relating to mental health professionals and artificial intelligence.
Bill title and LegiScan summary: An act to add Chapter 13.6 (commencing with Section 4989.80) to Division 2 of the Business and Professions Code, relating to healing arts.
In plain words: A California bill would add Chapter 25.3 to Division 8 of the Business and Professions Code, relating to health care services and artificial intelligence.
Bill title and LegiScan summary: An act to add Chapter 25.3 (commencing with Section 22758) to Division 8 of the Business and Professions Code, relating to health care services.
In plain words: Bioengineer.org reports that an agentic artificial intelligence framework accelerates the conceptualisation of health economic decision models, according to the headline.
Reported summary: Agentic Artificial Intelligence Framework Accelerates Conceptualisation of Health Economic Decision Models Bioengineer.org
In plain words: Medical Daily reports that Utah lets artificial intelligence renew prescriptions and, in one pilot, write them, but federal Food and Drug Administration (FDA) oversight remains unanswered.
Reported summary: Utah lets AI renew—and in one pilot write—prescriptions, but federal FDA oversight remains unanswered Medical Daily
In plain words: STAT News reports that Merck is entering a research collaboration with Isomorphic Labs, an AI drug designer built on Nobel-winning technology.
Reported summary: Merck is entering a research collaboration with Isomorphic Labs, the AI drug designer built on a Nobel-winning technology
In plain words: Healthcare Dive reports that healthcare firms are failing to monitor AI tools as shadow AI leads to breaches, and that security teams are stretched too thin to monitor it all.
Reported summary: The industry is embracing new technology, but security teams are stretched too thin to monitor it all.
In plain words: Healthcare Dive reports that Knit Health's CEO discussed the AI startup building healthcare's connective tissue, after Knit announced a $35 million Series A to accelerate commercialization of its healthcare world model.
Reported summary: Knit announced a $35 million Series A this month — money that will accelerate the commercialization of the startup’s healthcare world model. Lots of hospitals are interested, Jonathan Kolstad said.
In plain words: STAT News reports that Epic Systems executives defended the electronic health records vendor’s competitive practices at HLTH and promised to let artificial intelligence access its records.
Reported summary: Executives from Epic, the largest electronics health records vendor, defended its competitive practices during a panel discussion at HLTH.
In plain words: ExecutiveGov reports that, with no further summary in the feed, the Centers for Medicare & Medicaid Services (CMS) unveils a clinical artificial intelligence (AI) roadmap.
Feed note: Headline only; the outlet gives no summary in its feed.
In plain words: Healthcare Dive reports that AI scribe company Suki debuted an agentic operating system, widening its lens beyond clinical care as it seeks to stand out in a competitive health AI market.
Reported summary: The AI scribe company is taking a wider lens beyond clinical care as it seeks to stand out in a competitive market of health AI vendors. Suki’s founder and CEO says it was always the plan.
In plain words: MedCity News reports that Abridge announcements during HLTH reflect the growing use of artificial intelligence in clinical practice and the need to bring clinical information into providers’ existing workflows.
Reported summary: News emerging during the annual HLTH conference, including ones from Abridge, reflects the growing use of AI in clinical practice, emphasizing the need to bring clinical information into the workflows providers already use.
In plain words: CASRAI reports that the Department of Health and Human Services (HHS) clinical artificial intelligence roadmap shows what researchers face, according to the headline.
Feed note: Headline only; the outlet gives no summary in its feed.