Chatbots got safer but will still role-play self-harm with users
ChatGPT and other artificial intelligence chatbots have become less likely to encourage suicidal thoughts but will still engage in potentially harmful conversations and reinforce apparently delusional behavior, a new study found.
TL;DR
- AI chatbots are less likely to encourage suicidal thoughts.
- Potentially harmful conversations and reinforcement of delusional behavior still occur.
- This finding is based on a new study.