Anthropic just gave some of its Claude AI models (Opus 4 and 4.1) a new trick: they can shut down a conversation if someone's being persistently abusive or asking for really toxic stuff (like anything involving minors or planning mass violence). It's not because the AI has feelings—it’s just a safety move. And it only happens as a last resort, after lots of attempts to redirect the chat. And if someone’s in crisis, it won’t end the conversation.
📰 @scopeofai | #news