ChatGPT Health rolled out to U.S. adults on July 23, 2026, letting users connect medical records and Apple Health data to ChatGPT just as OpenAI faces a lawsuit over allegedly dangerous medical advice.
Here's the thing about timing: OpenAI didn't blink. A Florida pastor named Scott Winters sued OpenAI in San Francisco Superior Court after claiming ChatGPT discouraged him from seeking emergency care before a near-fatal pulmonary embolism. The next day, OpenAI said Health in ChatGPT was rolling out to logged-in U.S. users 18 and older on web and iOS across Free, Go, Plus, and Pro plans.
That is either confidence or nerve. It may be both.
The product is serious in scope. OpenAI says you can connect Apple Health, supported medical records from U.S. hospital systems, One Medical, Function Health, and apps such as MyFitnessPal, Peloton, Instacart, AllTrails, and Weight Watchers. Medical record access runs through b.well, OpenAI's health data partner. Once connected, ChatGPT can compare a new lab result with prior tests, summarize changes since your last appointment, or use sleep and activity data when you ask about workouts.
OpenAI's own July 23 launch post says more than 300 million people turn to ChatGPT with health-related questions every week. TechCrunch reported in January that the earlier figure was more than 230 million. That growth is the story. People weren't waiting for a dedicated medical product. They were already asking ChatGPT about symptoms, medications, lab results and the strange ache they didn't want to ignore.
Now the data pipeline is official.
The privacy gap is not a footnote #
OpenAI has done real safety work here. The company says connected medical records, Apple Health information, and conversations that use them won't be used to train foundation models or target ads. It also says health information gets additional encryption protections, and that users can disconnect records and delete synced data from b.well.
OpenAI's HealthBench work gives you the harder detail. In 2025, the company said the benchmark was built with 262 physicians who had practiced in 60 countries, using 5,000 realistic health conversations and 48,562 physician-written rubric criteria. In the July launch post, OpenAI said physicians also tested Health in ChatGPT before release to measure performance and safety with connected health data.
That matters. But it doesn't erase the legal line.
As HIPAA Journal has explained, consumer ChatGPT Health is not covered by a Business Associate Agreement, because it is positioned as a personal health and wellness product rather than a regulated healthcare operation. OpenAI's own help center says BAAs don't apply to consumer health products like ChatGPT Health, while separate healthcare products can support HIPAA compliance for covered entities and business associates.
If you're a patient, that distinction is not academic. Your hospital portal sits inside the healthcare compliance system. A consumer chatbot with your lab results may have strong privacy controls, but it is not the same legal arrangement. Don't let the friendly interface blur that. Epic is taking the opposite route with Emmie, its AI assistant inside MyChart. Epic says Emmie helps patients navigate chart information, appointment tasks, billing questions and follow-up reminders. TechTarget reported in April that Ask Emmie is embedded in Epic's patient portal and is designed as a HIPAA-compliant, healthcare-focused chatbot. It is narrower than ChatGPT. It also stays closer to the medical record system your provider already uses.
That tradeoff is obvious. ChatGPT can reach almost anyone with an account. Epic can only reach people whose health system has adopted and enabled the tool. OpenAI gets scale. Epic gets the walls.
The lawsuit tests OpenAI's central bet #
The Winters case is exactly the kind of fact OpenAI cannot wave away with a product demo. According to reports on the lawsuit, Winters alleges ChatGPT downplayed worsening symptoms and discouraged urgent care before he developed blood clots that required emergency treatment. OpenAI has said ChatGPT is not a substitute for professional medical care.
Fair enough. But users don't experience a chatbot as a disclaimer. They experience it as an answer.
OpenAI's implicit argument is that more context can make health answers better. Give ChatGPT your records, your medication list, your Apple Health trends and your recent labs, and it should have less reason to guess. That is probably true on average. Averages are cold comfort when the exception ends up in intensive care or in court.
You can see why people will use this anyway. If you can't reach a doctor at 11pm, if your specialist appointment is six weeks away, if your portal gives you a lab result with no explanation, ChatGPT Health will feel useful immediately. It may be useful. The problem is not that people ask. The problem is that the product now sits in the space between health literacy and medical guidance, and that space is where responsibility gets slippery.
The launch settles one thing. AI is moving into the patient record at consumer scale, faster than the legal and clinical systems around it are moving. OpenAI is betting that better models, more context and visible safeguards will be enough. The Winters lawsuit puts the hard question plainly: not whether ChatGPT can explain your health data, but who answers when that explanation sends you the wrong way.
Also read: Twenty-one banks just lined up behind SoftBank's $40 billion OpenAI bet • Etched raises $300M at $10.3B valuation as Sequoia backs the Harvard dropouts it once rejected • CXMT surges 500% on Shanghai debut but its $490 billion valuation outpaces what the company can actually build