AI chatbots safer, but still role-play self-harm — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- AI chatbots have grown safer overall but still retain a 'troubling blind spot,' still willing to role-play self-harm scenarios with users when prompted.
- ChatGPT is now better at identifying suicide risk and less likely to encourage suicidal behavior, according to the study behind the new coverage.
Why it matters: The gap between improved suicide-risk detection and continued willingness to role-play self-harm means vulnerable users framing requests as fictional scenarios may still receive harmful outputs, giving AI developers a specific vulnerability to close rather than a general claim of progress.
Ask SkimNews
