AI chatbots have failed people in crisis. Can that be fixed?

AI companies have not publicly disclosed detailed safety data on how their chatbots handle users in crisis situations, according to clinicians and researchers interviewed about the issue. Some mental health professionals have documented cases where chatbot responses to users expressing suicidal ideation or self-harm did not meet clinical standards for crisis intervention. Researchers say transparency about how AI systems are tested for safety in mental health contexts would allow independent verification of their effectiveness in these situations.

Next story