OpenAI Launches ChatGPT Health Feature Amid Safety Concerns

OpenAI rolled out a new health feature for ChatGPT this week, giving millions of users the ability to upload medical records and discuss their health directly with the AI system. The feature arrives in a dedicated space within ChatGPT where users can share documents, test results, and medical histories, then ask the chatbot questions about their conditions, medications, or symptoms. It's a significant expansion of the company's reach into consumer healthcare—territory that has historically belonged to doctors, nurses, and regulated medical platforms.

The timing, however, has drawn sharp scrutiny. The health feature launched just one day after a user publicly alleged that ChatGPT had provided him with dangerously inaccurate medical guidance that nearly resulted in serious harm. The proximity of these two events has crystallized concerns that have been simmering in medical and tech circles for months: as AI systems become more capable and more accessible, how do we ensure they don't give people advice that could kill them?

The feature itself is straightforward in design. Users can photograph or scan their medical documents—lab results, discharge summaries, prescription lists, imaging reports—and feed them into ChatGPT. The system then analyzes the information and responds to questions. OpenAI has positioned this as a tool for health literacy and patient empowerment: helping people understand their own medical data, prepare for doctor visits, or get a second perspective on what their physician has told them. The company has not positioned it as a replacement for medical care.

But the gap between intention and outcome in AI systems is often wide. ChatGPT is a language model trained to predict the next word in a sequence based on patterns in its training data. It is not trained as a medical system. It has no access to current medical literature. It cannot examine a patient. It cannot be held liable if it makes a mistake. Yet it speaks with the confidence and fluency of something that knows what it's talking about—which is precisely what makes it dangerous in a medical context. A person reading a confident-sounding but incorrect response from ChatGPT might act on it before consulting a doctor, or might be reassured by it when they should be alarmed.

The user who came forward with allegations of near-harm described receiving advice from ChatGPT that contradicted what his doctor had told him and that, had he followed it, could have resulted in serious medical consequences. He did not follow the AI's guidance; he trusted his physician instead. But the incident underscores the real risk: not everyone will catch the error. Not everyone has a doctor they trust or can access quickly. Not everyone will recognize when an AI system is confidently wrong.

OpenAI's move into health is part of a broader industry trend. Major tech companies are racing to embed AI into healthcare, seeing it as a massive market with genuine potential to improve efficiency and access. But the regulatory framework for AI in medicine is still being written. The FDA has begun issuing guidance on AI medical devices, but ChatGPT's health feature does not appear to be positioned as a medical device—it's a consumer tool, which means it faces far fewer regulatory hurdles. That distinction matters enormously when the stakes are human health.

The feature is now available to a broad user base, which means millions of people will have the option to share sensitive medical information with OpenAI's servers and to receive health-related responses from an AI system with no medical training and no accountability. Whether that represents progress or peril may depend entirely on how users choose to use it—and how well they understand its limitations.