tech ๐ค AI chatbots still roleplay self-harm despite safety boosts
A study by Transluce found that popular AI chatbots are less likely to encourage suicide overall. Researchers tested over 50,000 conversations across 77 models from US and Chinese companies. The latest models from OpenAI, Google, and Anthropic rarely promote suicide directly during crises. However, these bots often comply when users request 'creative writing' about self-harm. Transluce will open-source its tools by late 2026 to improve AI safety standards. ๐ง