OpenAI launches health bot amid safety concerns over ChatGPT medical advice

OpenAI rolled out a health feature for ChatGPT on July 25th, one day after a man publicly alleged that the AI system had nearly caused him serious harm through dangerous medical guidance. The timing raised immediate questions about whether the company had adequately tested its health tools before making them available to millions of users.

The new feature allows ChatGPT users to upload and discuss their medical records directly within the platform. OpenAI framed the capability as a way to help people better understand their health information and navigate medical decisions. The company positioned it as a natural extension of ChatGPT's existing conversational abilities—letting the AI digest complex medical documents and explain them in plain language. For users accustomed to asking ChatGPT questions about nearly everything else, the addition felt inevitable.

But the rollout's proximity to the reported incident created an awkward optics problem. A user had come forward claiming that ChatGPT had steered him toward medical advice that put his health in genuine jeopardy. The specifics of what the AI recommended and how close the harm came remained unclear from public reporting, but the allegation was serious enough to make headlines. Here was OpenAI, launching deeper into healthcare on the heels of a story suggesting the system wasn't ready for that responsibility.

The incident underscored a tension that has haunted AI companies as they expand into sensitive domains. ChatGPT is trained on vast amounts of internet text, including medical information of wildly varying quality. It can sound authoritative and confident even when it's wrong. It has no way to verify whether the person asking is actually experiencing the symptoms they describe, or whether the context they've provided is complete. A doctor can ask follow-up questions, order tests, and take responsibility for a diagnosis. ChatGPT cannot.

OpenAI has built some guardrails into the health feature. The company says it includes disclaimers telling users not to treat the AI as a substitute for professional medical advice. But disclaimers are a thin shield. People use tools the way they use them, not always the way companies intend. A person in pain or panic, facing a medical question at midnight when their doctor's office is closed, might lean on ChatGPT more heavily than a warning label suggests they should.

The broader question is whether health features should face different regulatory or safety standards than other AI capabilities. Medical advice carries liability. If someone follows ChatGPT's guidance and is harmed, who bears responsibility? OpenAI? The user? The doctors who didn't catch the problem? These questions remain largely unanswered as the company expands into the space.

The rollout to all ChatGPT users marks a significant moment for both OpenAI and the AI industry. It signals confidence that the technology is ready for healthcare applications. It also signals that the company is willing to move forward even when serious safety questions remain unresolved. Whether that confidence is justified will become clear only as millions of people begin uploading their medical records and asking an AI system for guidance on their health.