Anthropic bans cruelty to Claude from Nov 12

From the 12th of November 2026, repeated, needless cruelty toward Anthropic's Claude models breaks the company's rules. Anthropic announced the change in a blog post on the 8th of October 2026.

The new line bans "sustained and needless abusive or cruel behavior toward our models", Anthropic's Usage Policy page says.

Ending the chat stays the main penalty, Anthropic said

"Claude's ability to end these interactions will remain the primary enforcement mechanism," Anthropic wrote. Claude can already end rare abusive chats on Claude.ai and Claude Code, the post says.

Anthropic first gave this power to Claude Opus 4 and 4.1 on the 15th of August 2025, its post that day says. Claude uses it only after many tries to steer the chat fail.

Below we break down what changes on the 12th of November

Until now, Claude could end an abusive chat. From the 12th of November, the abuse also breaks the rules, the October post says.

The ban covers only extreme cases, where users "repeatedly act cruelly" with "no discernible purpose", the post says. Frustration, pushback, dark fiction and testing stay allowed.

The policy covers all users of Anthropic's products, its policy page says. That includes apps, the API (how developers plug Claude into software) and products built on it.

What users see when Claude ends a chat

The user can send no new messages in that chat, the August 2025 post says. Other chats stay open, and a new one can start at once.

What is confirmed and what is still open

For any violation, Anthropic "may warn you or throttle, limit, suspend, or terminate your access", its policy page says. The October post does not say if abuse leads there.

The post names chat ending only on Claude.ai and Claude Code. Neither Anthropic page says how the rule works on the API.

CBS News reported on the 9th of October that Anthropic did not immediately say what counts as "abusive" or "cruel".

Key facts

Announced by Anthropic on the 8th of October 2026.

In force from the 12th of November 2026, 35 days after the announcement, by our arithmetic.

Bans repeated, needless cruelty to Claude.

Claude ending the chat stays the main penalty.

Frustration, pushback, dark fiction and testing stay allowed.