AI-Generated Diet Advice Triggers Hospitalization and Highlights Transparency Gaps

A 60-year-old man sought to lower his dietary chloride by asking a chatbot for alternatives to table salt. The AI suggested sodium bromide, a chemical compound he substituted for salt. After several days he experienced hallucinations and paranoia, prompting emergency care. Clinicians could not retrieve the original chatbot exchange, leaving them unable to confirm the exact advice given. The incident is documented in a paper co-authored by researchers from Binghamton University, Stanford University, Texas A&M University and Indiana University.

Newsweek

The authors argue that AI-generated health guidance often lacks visibility, making adverse outcomes difficult for clinicians, researchers and regulators to trace. A Gallup survey of more than 5,500 U.S. adults conducted between October 27 and December 22, 2025, found that about 14 million adults reported skipping a health-care visit after using AI. The study also notes that AI can present incorrect answers with confidence, oversimplify complex medical information, and embed health content in social-media feeds and search results, exposing users who are not actively seeking medical advice.

Newsweek

An OpenAI spokesperson emphasized that ChatGPT is not a doctor and should never replace professional diagnosis or treatment. The company reiterated its terms warning against medical use and said it is working with hundreds of physicians from 60 countries to improve the model’s ability to recognize urgent situations and handle uncertainty. OpenAI also highlighted ongoing efforts to make outputs more reliable and safe.

Newsweek

The paper proposes several safeguards: granting users access to full records of health-related chatbot conversations, establishing mechanisms for users to flag and report harmful advice, and enabling independent third-party monitoring of AI health outputs. The authors further suggest that policymakers consider extending physician malpractice liability to AI chatbot providers and that social-media platforms improve labeling of AI-generated health content.

Newsweek

• “I think AI can be very helpful for health questions,” — Kaicheng Yang, one of the paper's authors

Newsweek

• “AI companies should give users access to records of their health conversations and a way to report harm, and independent third parties should be able to monitor and evaluate the health advice these systems give," Yang explained.” — Kaicheng Yang, one of the paper's authors

Newsweek

• “We do not think this is something we should rely on the companies to do, because their incentive is always to make more money," Yang said.” — Kaicheng Yang, one of the paper's authors

Newsweek

• “Sometimes I'm not even looking for health information, but just by browsing my social media feeds, it's there. It just shows up, and we believe that could have undesirable outcomes, especially if there's medical misinformation or state actors trying to manipulate the online discussion,” — Kaicheng Yang, one of the paper's authors

Newsweek