Dr. ChatGPT will see you now

Welcome to The Download, where we give you the day’s biggest news in bite-sized, context-filled pieces.

TL;DR: There’s a new, dedicated Health mode in ChatGPT, formalizing what users were already doing: asking it medical questions. OpenAI promises stronger privacy protections and more tailored health guidance—but the update blurs the line between offering context and dispensing potentially dangerous pseudo-medical advice.

What happened: Over 230 million people turn to OpenAI’s chatbot for medical advice each week (ChatGPT, what exactly is this rash?). With the ChatGPT Health update, OpenAI says users can “securely” share their health information with the chatbot. They can also now connect medical records directly to ChatGPT, link their wearables, and pull data from popular health and fitness apps like Apple Health, MyFitnessPal, Peloton, and even AllTrails.

Per OpenAI, conversations in Health are encrypted, isolated from your main chat history, and not used for model training. The company says physicians helped shape its safety guidelines and that ChatGPT should redirect users to real medical professionals when appropriate—though it’s not yet clear where that line is drawn. OpenAI says the tool isn’t for diagnosis or treatment—like that’s going to stop people.

Why it’s a big deal: It’s no surprise that OpenAI, which has seen stalling user growth for its core product, would roll this out. If people are going to ask ChatGPT about confusing lab results or medical questions, the tool should at least be designed to respond in a personalized, safer way. Health mode is also meant to increase accessibility: A user can ask medical questions at all hours, when many clinics are closed, regardless of insurance.

The bad: Say it with us: pri-va-cy. Users already hand over an alarming amount of personal information to ChatGPT, despite warnings from experts. An official health integration could lull users into thinking their medical data is completely secure and that ChatGPT has reliable answers to medical problems (and it’s worth noting that ChatGPT is not HIPAA-compliant).

As a reminder, ChatGPT’s track record isn’t spotless. Recently, a teen in California relied on the chatbot for drug-use guidance and ultimately died from an overdose. In another case, a man asked ChatGPT how to cut sodium out of his diet—after consulting it, he replaced table salt with toxic sodium bromide. He was later hospitalized.

Also notably absent from OpenAI’s announcement: any mention of mental health, an area where the company has faced intense criticism for its failure to do more to protect users showing signs of mental distress.

What to watch: OpenAI says it plans to add more health features down the road, making it even easier to give ChatGPT your entire medical history. —WK