Senator Murray Presses OpenAI on ChatGPT Medical Safeguards After Seattle Woman's Death

In a letter sent on Thursday, the Washington Democrat said it was unacceptable that chatbots are helping to guide potentially life-or-death decisions. "A chatbot is not a doctor," she wrote, saying no AI tool should steer people away from their physicians, and she called on OpenAI to "take accountability and provide assurance this can never happen again." The company has until December 1 to reply.

The letter adds congressional pressure to a lawsuit that already tests how far AI companies are responsible for health advice their products give. A West Health-Gallup Center on Healthcare in America report found that 25% of Americans have used AI chatbots for health information, and chatbots from Google, Microsoft and Anthropic are also used for medical questions. Murray told The Seattle Times that "this isn't an isolated incident," noting that chatbots are not doctors or nurses and are not reviewed by the Food and Drug Administration for safety and effectiveness.

Murray asked what safeguards OpenAI has added since the death of Bowen Wan, whether its systems flag when users report urgent symptoms, who reviews those responses, why the GPT-4o model was not suspended sooner, and whether the protections work in languages other than English.

Wan's parents sued OpenAI in October. Their complaint alleges that ChatGPT helped design a plan for her to taper off her anti-seizure medication and encouraged her to continue without telling her doctor, despite her repeatedly voicing reservations. It describes design features of GPT-4o as "uniquely dangerous" and accuses OpenAI of abandoning safety in the AI race. A medical examiner listed her cause of death as sudden unexpected death in epilepsy. The claims have not been tested in court.

GPT-4o had been criticised for being overly agreeable with users. OpenAI removed it from its consumer app on February 13, 2026, a day after the medical examiner believes Wan died, according to The New York Times, which also reported that Wan switched between English and Mandarin and sent an average of 1,500 messages a week to the chatbot in 2025.

In an emailed statement, an OpenAI spokesperson called the case "incredibly heartbreaking" and said the company is working with clinicians and other experts to adjust how ChatGPT responds in sensitive situations. The company said AI can help people make sense of health information but is not a substitute for medical care.

OpenAI's written answers to Murray's questions are due on December 1.