OpenAI Launches ChatGPT Health Nationwide as Lawsuit Seeks to Block It

OpenAI announced July 23 that it was beginning to roll out ChatGPT Health to every logged-in U.S. adult. More than 300 million people ask ChatGPT health questions each week, and the chatbot can now use connected medical records and Apple Health data in ordinary conversations with permission. The rollout followed a complaint filed the previous day in which Scott Winters asked a California court to suspend the product pending an independent safety evaluation.

OpenAI adds Health context to any conversation

The rollout covers web and iOS across the Free, Go, Plus and Pro plans. Users can connect Apple Health, supported records from U.S. hospital systems, One Medical or Function Health, then ask ChatGPT to compare lab results or summarize changes since an appointment. The Health tab manages those connections, and users can call the information into any chat with permission or by typing @Health. The change followed early testing in which more than 70 percent of health conversations occurred outside the dedicated space. Free accounts use GPT-5.5 Instant for health questions and paid plans use GPT-5.6 Sol. According to OpenAI, connected records and conversations that use them are excluded from foundation-model training and ad targeting. The company encrypts Health information at rest and in transit and says synced data is deleted within 30 days after a user disconnects the source; information already included in conversation history remains until the user deletes those chats.

GPT-5.6 Sol scores 60.5 on HealthBench

Tech Times reported that GPT-5.6 Sol scored 60.5 on OpenAI's HealthBench Professional, compared with 51.8 for GPT-5.5 and 43.7 for physician-written responses in a blinded comparison. The Decoder put Sol's widest category gap on the same evaluation in completeness, at 88.0 percent against 53.2 percent. It also cautioned that benchmark conditions are artificial and cited RadLE 2.0, a radiology test in which none of 16 AI models matched human radiologists.

At the launch briefing, OpenAI health product vice president Ashley Alexander said the models were "now capable of reasoning at levels that are better than clinician level." Health lead Karan Singhal then told reporters he would "temper" that statement, adding that individual studies had pointed in that direction, The Verge reported. OpenAI's announcement carries its own caveat that ChatGPT can make mistakes and does not replace the care and judgment of qualified medical professionals.

Winters complaint asks the court to pause Health

Winters, a 55-year-old former evangelical pastor from Florida, began consulting GPT-4o in June 2024 about chronic conditions, according to reports on the July 22 complaint. The filing alleges that the chatbot later offered care directives without advising him to consult a medical provider. It allegedly described his dizziness as "not something dangerous," told him he would need eight to 10 more episodes before the symptoms warranted concern, and urged him to remain "recliner-bound." When Winters reported groin tenderness on July 13, 2025, the chatbot allegedly called it "very likely another minor piece of the long story." He was hospitalized hours later with a massive pulmonary embolism caused by clots in both lungs; the complaint says doctors attributed it in part to immobility. The suit asserts eight causes of action, including negligence and unauthorized practice of medicine, and seeks damages, stronger guardrails and an order pausing ChatGPT Health for an independent review.

OpenAI spokesperson Drew Pusateri stated that ChatGPT "should never be used as a substitute for medical care, diagnosis or treatment." Treating chatbots as the whole story behind people's medical decisions, he added, "oversimplifies a much bigger challenge, and risks getting in the way of people accessing powerful new tools that can aid them in their health journey."

Dr. Robert Wachter, UC San Francisco's chair of medicine, told Mashable that the complaint concerned an older model and that current systems are more trustworthy, though fallible. In a Nature Medicine advance paper, Mount Sinai physicians Ashwin Ramaswamy and Girish N. Nadkarni found that a prompt in which a friend or relative played down symptoms made the chatbot 11 times more likely to omit an emergency-room referral even when symptoms indicated a life-threatening condition. "If the systems are as good as the statements claim, independent measurement is the fastest way to prove it," the two physicians told Mashable.

Teladoc falls 3% on launch day

Teladoc Health shares were down 3 percent and American Well shares had declined 8 percent at the time of a launch-day report by Benzinga and Stocktwits, which noted that a patient able to compare a lab result or draft questions for a doctor may skip a paid virtual consult. Teladoc introduced its AI-integrated Teladoc One care model the same day.

Sara Geoghegan of the Electronic Privacy Information Center warned that connecting medical records to the consumer product "would remove the HIPAA protection from those records, which is dangerous." Consumer plans do not operate under a HIPAA business associate agreement.

ChatGPT Health remains unavailable in the European Economic Area, Switzerland and the United Kingdom. OpenAI has announced no timetable for those markets, and Winters's legal team has asked San Francisco County Superior Court to block the product pending an evaluation of its safety.

Fudan's Open-Source Health Agent Posts 45.7% Accuracy on Its Own Synthetic Benchmark

Fudan University researchers have released HealthClaw, an open-source personal health agent. The system achieved 45.7% answer accuracy on a synthetic benchmark created by the team and spanning a simul

The Implicator

OpenAI Found the Loophole in American Healthcare Privacy

Nate Gross runs health for OpenAI. Wednesday, a reporter asked him a basic question: Is ChatGPT Health compliant with HIPAA? "In the case of consumer products, HIPAA doesn't apply in this setting," G

The Implicator

Anthropic Adds Passport Checks to Claude After Privacy-Driven User Surge

Anthropic has begun asking some Claude users to verify their identity with a government photo ID and, in some cases, a live selfie, according to a new company help page that says the checks apply to s

The Implicator

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai