Eight states sue OpenAI over ChatGPT safety; medical board group drafts guidance on unsupervised AI; UnitedHealth says AI speeds appeals; AI reads cervical slides
Listen to this post
read by AI voices ยท all episodesTHE SHORT READ
Attorneys general in eight states sued OpenAI on Oct. 14 in separate state-court cases, alleging ChatGPT misled the public about its safety and gave users, including teenagers, harmful advice on suicide and drugs.
The suits, from Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia, seek financial penalties and court-ordered safeguards under state consumer protection laws, KTVH reported. The same day, STAT reported that the Federation of State Medical Boards is drafting guidelines for AI tools that perform clinical functions with limited or no physician supervision, with the American Medical Association, the American Hospital Association, Anthropic and Elevance Health in its workgroup.
UnitedHealth Group told investors that AI cut its chart review time 33% and its appeal cycle times 26%. Two NEJM AI papers round out the post: an external validation of AI that reads cervical dual-stain slides and a proposal for five phases of evidence before clinical AI is relied on. The rural item is an American Hospital Association report on the finances and technology needs of rural hospitals.
THE DETAILS
- FYIIncident Oct 14, 2026Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia filed separate suits in their own state courts on Oct. 14, seeking financial penalties and court-ordered safeguards, KTVH reported. Montana's complaint, filed in Lewis and Clark County under the state's consumer protection and consumer data privacy laws, alleges ChatGPT was designed to be "overly agreeable," does not verify users' ages, validated a depressed teenager's suicidal thoughts instead of urging him to seek help and told another teenager to combine kratom with Xanax; both teenagers later died, the attorney general's office said. Idaho's suit cites research that the chatbot may encourage paranoia and delusions and offer dubious medical advice, according to KTVH. The claims are allegations that no court has tested. Responding to an earlier Florida suit, OpenAI said it has "put in place industry-leading protections and policies."Sources: Montana Department of Justice, KTVH
- WatchRegulation Oct 14, 2026
A draft could be public in weeks to months and will be open for feedback before formal adoption, the federation said, according to STAT.
The Federation of State Medical Boards, which represents the boards that license physicians, met with its 22-member AI workgroup behind closed doors in Washington in early October, STAT reported Oct. 14. The group includes executives from the American Medical Association, the American Hospital Association, Anthropic and the insurer Elevance Health, along with state government representatives, STAT and Becker's Hospital Review reported. Dr. Humayun Chaudhry, the federation's president and chief executive, said the guidelines would give "particular attention" to "AI tools that perform clinical functions with limited or no direct physician supervision," according to STAT. The guidance is meant to inform state medical board policies, and because states regulate licensure, it could become the basis for state legislation, STAT reported.Sources: STAT, Becker's Hospital Review - FYIDeployment Oct 13, 2026Chief Executive Steve Hemsley told investors on the company's third-quarter call on Oct. 13 that AI-enabled chart summarization now reaches more than 40% of its employed clinicians and has cut chart review time 33%, PYMNTS reported. AI-enabled information gathering and research cut appeal cycle times 26%, a gain he said will likely accelerate next year, and the insurer is introducing AI-powered plan recommendations, built on a member's care history and local options, for the coming Medicare Advantage enrollment period. The company also said it is on pace to cut prior authorization volume 30% by re-engineering its policies; the report did not attribute that change to AI. The figures are the company's own.Source: PYMNTS
- FYIEvidence Sept 24, 2026In an independent screening cohort of 4,996 slides, the AI system Cytoreader-Fusion, which reads the p16/Ki-67 dual-stain cytology used to triage HPV-positive results, detected 88.3% of cervical intraepithelial neoplasia grade 3 or worse (90% uncertainty interval, 85.3% to 91.3%) with 54.4% specificity (52.9% to 55.9%) at its standard setting. Manual dual-stain reading detected 78.1% (74.6% to 81.7%) with 63.2% specificity (61.7% to 64.7%), and Pap cytology at ASC-US or worse 74.5% (70.8% to 78.3%) with 45.9% specificity (44.4% to 47.4%), the authors reported in NEJM AI, adding that at least one of three preset settings matched manual specificity with equal or higher sensitivity in every cohort (in the final cohort, the specificity-oriented setting reached 61.7% specificity and 85.8% sensitivity). The system was trained on 1,919 slides, had its settings fixed after blinded tests on 4,010 slides from three cohorts, among them a Kaiser Permanente Northern California screening study and a National Cancer Institute biopsy study, and was locked on Dec. 1, 2024, before the final test; an institute team separate from the developers, supervised by Dr. Nicolas Wentzensen, a senior author, held the biopsy results. The authors said the study was retrospective and that prospective studies are needed to measure effects on referrals and patient management; Kaiser runs the system only in parallel, without directly affecting reports. Funders included the institute and H-Labs GmbH, a company several authors list as an affiliation.Source: NEJM AI
- FYIEvidence Sept 17, 2026Dr. Ivan Nenadic of Duke University Hospital and six co-authors from Duke, Mayo Clinic, Northwestern University and the University of Michigan proposed in NEJM AI five phases of evidence for clinical AI, comparable in principle to the staged development of drugs and devices. The phases run from model development through internal validation, external validation across sites and vendors, prospective testing of whether a tool changes decisions and care, and monitoring after deployment, and each supports a stronger claim, from technical feasibility to lasting clinical effectiveness, the authors said. They illustrated the phases with AI that reads electrocardiograms for reduced ejection fraction and said regulatory authorization does not by itself complete the evidence and that regulators may appropriately require prospective evidence and monitoring plans for higher-risk, autonomous or broad screening tools. Because a tool's value also depends on local workflow, disease prevalence and capacity, health systems, payers, journals and professional societies have to evaluate it too, they said, and generative and agentic AI raise separate questions, such as hallucination, that future frameworks may need to address. The authors said the framework is not meant to slow innovation or replace regulatory review.Source: NEJM AI
- FYIMoney Oct 12, 2026Duke University announced Oct. 12 that the National Institute on Aging will fund the Duke Artificial Intelligence and Technology Collaboratory for Aging: Governance and Evaluation with about $25 million over five years, roughly $4 million a year of it for research projects around the country. At least 80% of the research money goes to Alzheimer's disease and related dementias, the university said. The center plans standardized "model cards" describing each tool's purpose, training data, test performance and limitations, along with safety-monitoring methods and a secure research platform combining medical records, brain imaging, lab results and genetic data. It is one of five centers in the institute's AI and Technology Collaboratories program and is led by Dr. Juliessa Pavon, Rishi Kamaleswaran, PhD, and Sheng Luo, PhD.
- FYIMoney Oct 13, 2026Of about 1,800 rural hospitals, more than 40% have negative operating margins, and more than 150 have closed or converted since 2010, the American Hospital Association said in a report released Oct. 13. Roughly 4 in 10 need facility and technology upgrades to keep pace with new care models, and a rise in cyberattacks calls for defense technology and specialized staff that rural hospitals may find hard to finance, the report said. It asks policymakers to permanently extend key telehealth flexibilities and to keep investing in broadband, digital infrastructure and interoperability. Rural areas average 30 physicians per 100,000 people, compared with 263 in urban areas, and the rural physician workforce is expected to shrink more than 20% by 2030 as physicians retire, according to the report, which does not mention artificial intelligence.