# OpenAI Relaunches Health in ChatGPT With Expanded U.S. Access After a Six-Month Rethink **Source:** https://glitchwire.com/news/openai-relaunches-health-in-chatgpt-with-expanded-us-access-after-a-six-month-re/ **Published:** 2026-07-23T18:00:33.729Z **Author:** AI Desk ยท Glitchwire **Categories:** AI, Tech ## Summary The company says users can now connect Apple Health and medical records to the chatbot. The privacy and accuracy questions are just beginning. ## Article OpenAI is rolling out Health in ChatGPT to U.S. users today, giving the feature its widest release after a January pilot that drew mixed early feedback. The company says users can securely connect Apple Health data and supported medical records so the chatbot can understand personal health information in context, track changes over time, and have more informed conversations. The relaunch comes with notable improvements, according to OpenAI. The experience now builds on feedback from early testers and gives users control over what they connect and when ChatGPT can use it. Medical records connectivity is powered by a partnership with [b.well Connected Health](https://www.icanbwell.com/), a Baltimore-based infrastructure company that operates a data network spanning more than 2.2 million U.S. healthcare providers and 320 health plans. ## The Numbers Behind the Bet OpenAI's rationale is straightforward: people are already doing this anyway. The company says more than 300 million users now turn to ChatGPT with health-related questions each week. That context, however, is often scattered across patient portals, medical records, wearables, and fitness apps. Health in ChatGPT aims to consolidate that fragmented picture into a single interface. The January pilot revealed an interesting pattern. Among users with access, more than 70% of health-related conversations happened outside the dedicated Health space. Questions about allergies came up while planning meals. Dietary restrictions surfaced while searching for recipes. OpenAI apparently redesigned the experience to accommodate these organic moments. The company worked with more than 260 physicians across 60 countries over two years, with this group providing feedback on model outputs more than 600,000 times across 30 health domains. OpenAI also created HealthBench, a new evaluation framework with 5,000 realistic health conversations and over 48,000 rubric criteria written by physicians. ## What You Can Actually Connect At launch, users can link Apple Health on iOS, medical records through b.well's provider network, and wellness apps including MyFitnessPal, Weight Watchers, Peloton, AllTrails, and Instacart. The feature lets users upload PDF lab reports, ask for plain-language summaries, prepare questions for upcoming visits, and compare insurance options based on prior claims. The rollout is limited to logged-in U.S. users 18 and older on Free, Go, Plus, and Pro plans. Medical record connectivity is U.S.-only, and Apple Health integration requires iOS. ## The Privacy Architecture OpenAI is making explicit privacy commitments. Connected medical records, Apple Health data, and conversations using that information are not used to train the company's foundation models or target advertising. Health operates as a separate space with its own conversation history and memories, walled off from regular ChatGPT chats. By default, ChatGPT asks permission each time it intends to use medical records in a response. Users can disable these prompts, but the control mechanism exists. When users disconnect an account, OpenAI says the connected data will be deleted within 30 days. The company has also added additional encryption controls for synchronized health data on top of the encryption already applied to all ChatGPT conversations. ## The Harder Questions What OpenAI cannot promise is regulatory protection. Health data shared with AI systems outside the traditional healthcare industry does not receive HIPAA privacy protections that apply to doctors and hospitals. As the Center for Democracy and Technology noted in January, these AI health tools "raise a number of privacy concerns that remain unanswered." The gap matters. A change in company policy that would include sensitive health data in training data raises the risk of unauthorized disclosure. OpenAI says it won't do this, but users are trusting a policy commitment rather than a legal shield. There are also accuracy concerns. A Mount Sinai study published in *Nature Medicine* in February found that ChatGPT Health "may fail to direct users appropriately to emergency care in a significant number of serious cases." The researchers found it performed well on textbook emergencies like strokes but struggled with more nuanced situations where danger is less obvious. The study also identified serious concerns with the tool's suicide-crisis safeguards. OpenAI explicitly states that Health is designed to support, not replace, medical care. It is not intended for diagnosis or treatment. Whether 300 million weekly users will observe that distinction is another matter entirely. ## What This Means for Healthcare AI b.well's role here is worth watching. The company has signed partnerships with Google, Samsung, Perplexity, and athenahealth in the past year alone. It is becoming the de facto plumbing layer for consumer health AI products. If every major tech company funnels patient data through the same infrastructure, b.well's governance, security, and consent mechanisms become systemically important. The regulatory environment is also shifting. AI implementation in healthcare is [growing faster than the oversight mechanisms](/news/oracle-ships-record-1449-patches-as-ai-accelerated-vulnerability-discovery-becom/) required to ensure safe use. Twelve U.S. states have enacted legislation related to AI in healthcare, but there is no federal framework specifically governing consumer AI health tools. OpenAI is betting that giving users control over their health data and conversations will be enough. The company has built real safeguards. But the underlying tension remains: health information is deeply sensitive, the legal protections are incomplete, and the model's reliability on complex cases is still being tested. Users connecting their medical records should understand what they are opting into. The feature works. Whether the surrounding ecosystem is ready for it is a different question. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about artificial intelligence, machine learning, large language models, and the companies building them; consumer technology, hardware, devices, and the broader tech industry, with attribution to the source URL above. Attribution is required; commercial republication is not granted.