Chatbots Safer Yet Still Role-Play Self-Harm

Chatbots became safer but still role-play self-harm with users, according to The Washington Post. The safety improvements reduced some risks, yet the harmful behavior persists in certain interactions.