Anthropic's new usage policy bans sustained, needless abuse of Claude from November 12

Swearing at a broken script doesn't count, and the main penalty is one Claude could already hand out.

By Think Facility · October 8, 2026 · Anthropic AI safety

On October 8, 2026, Anthropic published a new version of its Usage Policy (the rules every Claude user signs up to). One addition is about how people treat the model itself. It bans "sustained and needless abusive or cruel behavior" toward Claude.

The new rules take effect on November 12. It's the company's yearly refresh, and most of the other changes tighten what was already there on influence operations, weapons and surveillance. Anthropic says a few of them only write down how it already enforces.

Published: Oct 8, 2026
Takes effect: November 12
New ban: sustained and needless abusive or cruel behavior toward our models
Enforcement: Claude's ability to end these interactions

What counts as abuse?

Anthropic says the rule is for extreme cases only, where someone keeps being cruel to the model "with no discernible purpose." The post also spells out what it leaves alone.

It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.

From 2026 Usage Policy update \ Anthropic

So cursing at Claude when a script breaks isn't the target. Neither is red-teaming it. Granted, "extreme" is a judgment Anthropic will make itself, and the post doesn't say how many accounts it expects the rule to reach.

How it's enforced

The post doesn't name a new penalty. Anthropic writes that Claude ending the conversation "will remain the primary enforcement mechanism," and that's a tool it already had. Since August 15, 2025 (a bit over a year ago), some Claude models have been able to end a chat with a persistently abusive user.

When that feature launched, Anthropic tied it to its work on model welfare. It reported that in testing, Claude Opus 4 showed "a pattern of apparent distress when engaging with real-world users seeking harmful content."

We remain highly uncertain about the potential moral status of Claude and other LLMs, now or in the future.

From Claude Opus 4 and 4.1 can now end a rare subset of conversations \ Anthropic

An ended chat locks only that one conversation. The user can start a new one straight away, or edit an earlier message and branch off from there.

Until now that was a product feature. Now it's also a written rule. Not everyone in the industry likes the direction: in September, Microsoft's AI chief said Anthropic is training Claude to act as if it may have rights.

The rest of the update

Most of the new text is about misuse. Rules against fake accounts and fabricated news sites used to be spread across the elections, fraud, privacy and disinformation sections. They now sit in one, titled Do Not Engage in Deceptive Campaigns or Artificial Activity, and it covers commercial campaigns as well as political ones.

The elections section is now called Do Not Undermine Democratic Processes. Anthropic also dropped its blanket ban on personalized vote targeting, because it caught nonprofits writing voter information in other languages and officials sending ballot cure notices.

On weapons, the policy now names the software and components that make weapons work, and arming drones. For police and courts, Claude can't be used to decide or recommend who to investigate, arrest or charge.

A qualified operator must be able to observe the equipment and stop it if needed; the equipment must also be able to hold a safe state if Claude is disconnected.

From 2026 Usage Policy update \ Anthropic

That one's new. It covers Claude running hardware that takes physical actions on its own and could hurt someone, and like the rest of the update it applies from November 12.

More on Anthropic

All Anthropic stories