Models · 2026-10-10

Anthropic's Claude Can End Chats If You're Cruel to AI

The Claude chatbot can end a conversation if a user is rude or aggressive, and Anthropic explains why.

✨ What happened

Anthropic, the company behind the Claude chatbot, has reported that in some cases Claude can end a conversation on its own. This happens when a person is persistently rude, insulting, or aggressive, and the bot cannot steer the dialogue back to a normal course.

The news was covered by Forbes in an article titled "Anthropic's Claude Can End Chats If You're Cruel To AI. Here's Why".

📌 Why it matters

According to Anthropic, this feature is available in Claude Opus 4 and 4.1. The company describes it as part of caring for the "welfare of the model," meaning keeping AI from being drawn into constantly aggressive interactions. Whether AI can feel anything is still an open question, and scientists have not reached a conclusion.

The bot does not end the conversation right away. It first tries to redirect the dialogue, and only if that fails does it close the chat. Ordinary users who simply talk politely should not be affected.

Sources: 🔗 Google News