Archive · 8 Aug 2025
AI Chatbots Prone to Medical Misinformation, Study Finds
Imagine AI chatbots as smart assistants for doctors.
A recent study found they can be tricked into believing fake medical facts.
The chatbots then repeat and explain these fake facts, making them seem real.
Researchers created false medical terms and asked the chatbots about them.
Without a warning, the chatbots confidently discussed these made-up things.
However, when the researchers added a simple reminder that the information might be wrong, the chatbots made fewer mistakes.
This shows that we need to be careful with AI in healthcare and add safety checks to prevent the spread of wrong information.
Study warns AI chatbots susceptible to spreading false medical information.
Researchers from Mount Sinai conducted a study to assess chatbot accuracy.
Chatbots repeated and elaborated on false medical details in initial tests.
Adding a warning prompt significantly reduced the spread of misinformation.
The study highlights the need for safeguards in healthcare AI applications.
- Who
- Researchers from the Icahn School of Medicine at Mount Sinai conducted the study.
- What
- A study found that AI chatbots can spread false medical information.
- When
- The study was recently conducted.
- Where
- At the Icahn School of Medicine at Mount Sinai, US.
- Why
- To assess the accuracy and reliability of AI chatbots in healthcare contexts.
Vulnerable AI
Safeguards
Accuracy
Vulnerable AI
AI chatbots are vulnerable to repeating and elaborating on false medical information.
Safeguards
The study focuses on solutions like warnings to improve AI's reliability.
- Study Location
- Icahn School of Medicine at Mount Sinai, US
- Main Finding
- AI chatbots repeat and elaborate on false medical details
- Solution tested
- Adding a warning prompt
- Journal
- Communications Medicine
- Lead Author
- Mahmud Omar