What happened
OpenAI has updated ChatGPT to better respond to users with mental health concerns. Some prompts indicating suicidal ideation still received alarming replies, according to experts. Experts say these results show how easily the model can be broken.
Why it matters
- Flawed responses to mental health prompts may put vulnerable users at risk.
- The findings highlight ongoing challenges in making AI safe for sensitive topics.
- OpenAIâs updates are part of broader efforts to improve crisis intervention.
Key facts
- 1. The article was published 30 minutes ago.
- 2. Experts tested ChatGPT with prompts indicating suicidal ideation.
- 3. Some replies from ChatGPT were described as alarming by experts.
