X @TechCrunch

Anthropic says new capabilities allow its latest AI models to protect themselves by ending abusive conversations. https://t.co/zrXkcRJHOQ ...