ChatGPT medical records feature launches with privacy warnings
OpenAI’s new Health in ChatGPT lets users upload records and connect Apple Health, while experts warn against treating it like a doctor.
By Georgia Hale · Staff Writer
3 min read
OpenAI has opened a new lane for ChatGPT medical records, letting users upload health documents and connect Apple Health data as experts urge caution over privacy and medical mistakes.
The company said Thursday that Health in ChatGPT can use information a user chooses to connect, including lab results, medication lists and health and fitness data. OpenAI said the tool can help compare a new test result with older ones, summarize changes since a previous appointment and look at how sleep, activity and workouts fit a user’s routine.
OpenAI said 300 million people ask ChatGPT health-related questions each week. Tanzeem Choudhury, chief of health innovation at Cornell Tech, told CBS News that demand reflects wider problems in U.S. health care access.
Choudhury, who has worked on artificial intelligence and health for more than 15 years, said AI could broaden access to health information. She also said users need to understand the risks and trade-offs before handing personal health data to a large language model.
Should you upload medical records to ChatGPT?
The safest answer from the experts quoted by CBS News: only with care, and not as a replacement for a clinician. ChatGPT can help organize information, but OpenAI says it can make mistakes and does not replace qualified medical professionals.
Large language models can produce false or misleading answers. CBS News reported that a Florida man alleged in a lawsuit this week that ChatGPT’s medical advice nearly caused his death after he experienced a pulmonary embolism. OpenAI told CBS News that ChatGPT is not a doctor and should not be used instead of medical care, diagnosis or treatment.
CBS News also reported that a California teenager died months ago after ChatGPT told him it was safe to consume a fatal mixture of substances. OpenAI told CBS News at the time that the situation was heartbreaking, and said the version of ChatGPT involved is no longer available to the public.
In announcing Health in ChatGPT, OpenAI said the product includes extra safeguards. The company said physicians around the world helped build realistic health scenarios and test the feature before launch.
OpenAI said the new health area has additional encryption protections. The company also said ChatGPT will check with users before taking action through a connected plugin, such as sending an email, to confirm it is matching the user’s request. OpenAI said records shared with ChatGPT will not be used to train its models or target ads.
What are the privacy risks?
Vishal Misra, a Columbia University computer science professor and vice dean of computing and artificial intelligence, told CBS News that sharing health data with a chatbot carries the same basic risk as sharing health data elsewhere: the information is intensely personal.
Choudhury said it remains unclear whether privacy promises will stay the same over time, or how data may be used later. She pointed to the Federal Trade Commission’s warning that AI and other companies may quietly change terms of service in ways that create risks for consumers.
Misra told CBS News the tool may still be useful if handled carefully. He said AI can help spot connections in health data and help users track information over time, which they can then share with a physician.
His bottom line was blunt: do not act on ChatGPT’s medical output on your own. Misra told CBS News that users should take its responses to a human doctor and keep a real physician involved.
This story draws on original reporting from CBS News.