Science & Tech · AI · 1 day ago
Researchers warn harmful AI health advice can be hard to trace
People are using AI chatbots for health advice, but that advice can be wrong.
If someone is harmed, doctors may not be able to see the chatbot conversation that led to it.
That makes it harder to understand what happened and spot similar problems.
Researchers want companies to help users access chat histories and flag risky advice, and say policymakers should consider holding companies liable.
Researchers from Binghamton University, Stanford, Texas A&M and Indiana University published a paper in Nature Health about the risks of harmful health advice from AI chatbots.
The researchers say chatbot conversations may remain accessible only to users and AI companies, leaving doctors, regulators and independent investigators without evidence of what advice was given.
The paper describes a 60-year-old man who was told by ChatGPT to replace table salt with sodium bromide and was later hospitalized with bromide toxicity.
The researchers also warn that people can encounter harmful AI health information through summaries and social feeds without asking a chatbot a question.
They recommend giving users access to their chat histories, developing systems to flag harmful advice and considering whether AI companies providing medical guidance should face physician malpractice liability.
- Who
- Researchers from Binghamton University, Stanford, Texas A&M and Indiana University.
- What
- A paper examining the risks of harmful AI chatbot health advice that may leave no record for doctors or regulators.
- When
- The article was published on October 9, 2026; the paper's publication date is not stated.
- Where
- Not stated
- Why
- The researchers say unrecorded chatbot conversations can make harmful advice difficult to identify and investigate.
This story does not have two clearly opposing sides.
No direct quotes in the coverage so far.
This story does not have a timeline yet.
- Paper's journal
- Nature Health
- Researchers' institutions
- Binghamton University, Stanford, Texas A&M and Indiana University
- Man in the example
- 60 years old
- Advice described
- ChatGPT told him to replace table salt with sodium bromide.
- Reported outcome
- He was hospitalized with bromide toxicity, hallucinations and paranoia.
Sources
Related news
Nadella says advanced AI needs human-controlled emergency brake
Anthropic AI submits false murder tip to Philadelphia police
Anthropic bars repeated cruelty toward its AI models
OpenAI fires three safety researchers after dispute over sensitive information
Gemini in Chrome can summarize pages and compare information across tabs
Major tech figures adopt Trump’s preferred term for AI
Nikon disqualifies AI-assisted microscope video and names a new winner
Microsoft adds security tools and local AI features for Windows
Acemoglu says AI could triple unemployment within a decade
UK privacy watchdog secures data protection commitments from 10 AI companies
Anthropic cuts internet access for all internal AI evaluations
