On Friday, Anthropic said its Claude chatbot can now end potentially harmful conversations, which "is intended for use in ...