OpenAI has launched a new feature that connects ChatGPT directly to users' medical records and fitness trackers, sparking fresh concerns about data privacy and the safety of AI-generated health advice. While the tool is technically restricted to adults, its release highlights the urgent need for parents and educators to teach teenagers how to use medical AI safely. As we previously reported, preparing students for an AI-driven world requires proactive boundaries and real AI literacy.
What Happened
According to Fierce Healthcare, the new "Health in ChatGPT" feature is rolling out to logged-in U.S. users aged 18 and older across all subscription tiers, including free accounts. Users can link their personal health profiles, syncing data from platforms like Apple Health and clinical record systems like Epic or One Medical. OpenAI claims this feature helps users organize scattered records, allowing the chatbot to consider medical history and past injuries during everyday chats. OpenAI states that these health-related conversations are stored with extra encryption and are not used to train its AI models or target advertisements.
The Bigger Picture
Despite these promises, privacy and safety experts are raising alarm bells. Connecting medical records to a consumer app permanently strips those files of federal HIPAA protections, leaving users governed solely by OpenAI's terms of service.
The medical advice provided by consumer AI also faces severe legal challenges. A California family recently filed a wrongful death lawsuit after their teenager died from consuming a fatal combination of substances that ChatGPT had falsely told him was safe. In another active lawsuit, a Florida pastor alleges he nearly died from a pulmonary embolism after the chatbot dismissed his symptoms and told him not to seek emergency medical care.
Clinical studies confirm that these models struggle with accuracy. A study published in Scientific Reports found that nearly 4% of ChatGPT’s answers to pediatric neurology questions were completely incorrect, and over 31% were incomplete. Another study published in the medical literature showed that while AI could accurately diagnose certain common rashes, its diagnostic accuracy dropped to 0% for insect bites. This shows that AI struggles with complex, real-world context.
What This Means for Families
Even though OpenAI restricts this feature to adults, teenagers regularly bypass age gates or use shared family accounts. Families who share a single ChatGPT Plus subscription run the risk of exposing sensitive medical records to children. Teens who use ChatGPT as a homework helper may also use it as an unsupervised school nurse, asking it about puberty or mental health. Without active guidance, young users may mistake the chatbot's confident, conversational tone for clinical expertise.
What You Can Do
To protect your family, start by opening the settings menu of any shared ChatGPT account and ensuring no personal health records or Apple Health databases are connected.
Next, talk to your children about clinical skepticism. Share the research showing that AI models make critical diagnostic errors and explain that a chatbot's confident tone does not mean its medical advice is accurate.
Finally, establish a strict rule that any physical symptoms or medication questions must go to a parent, school nurse, or pediatrician rather than a screen.