Anthropic bans abusive behavior toward Claude starting November 12

Anthropic has published a revised Usage Policy for Claude. The new version was announced on October 8 and takes effect on November 12, 2026. It applies to everyone who uses Claude, including people on the Claude apps, on Claude.ai and on the free plan, which recently picked up Claude Haiku 5.5. According to Anthropic, most of the changes clarify existing rules. One of them is new and concerns how users treat the model itself.

What counts as abuse

The policy now bars users from engaging "in sustained and needless abusive or cruel behavior toward our models." In its announcement of the update, Anthropic says the rule targets extreme cases only. Ordinary frustration, pushback, dark creative themes and model testing or research are not covered.

Enforcement relies mostly on a feature Claude already has. Since August 2025, some Claude models have been able to end a conversation when a user stays abusive after repeated attempts to redirect it. Anthropic introduced the feature as part of its research on model welfare. Once a chat has been ended, no new messages can be sent in it, but other conversations are unaffected and a new chat can be started right away. "Claude's ability to end these interactions will remain the primary enforcement mechanism," the company writes. The general enforcement clause of the policy still allows Anthropic to "warn you or throttle, limit, suspend, or terminate your access" if it suspects a violation.

Tighter rules on surveillance and weapons

The surveillance rules are now more specific. Claude may not be used to track a person without their consent, either in real time or by analyzing data collected earlier. Using Claude to decide or recommend who should be investigated, arrested or charged is also banned. Fraud monitoring that people have agreed to, content moderation, journalism and legal research remain allowed.

On elections, Anthropic removed the blanket ban on personalized vote and campaign targeting. Targeting that relies on deception or misuses voters' personal data is still prohibited under other sections. A new section brings together the rules against networks of fake accounts and fabricated news sites, which Anthropic says it has seen in recent misuse. The weapons ban now explicitly covers the software and components that make weapons work, including systems used to arm drones.

What changes for you

For most people, very little. The rule only applies to sustained abuse, so anyone using Claude for writing, research or code is unlikely to notice it. The policy does not say whether your chats are used to train future models. That is controlled in Claude's privacy settings under Model Improvement. Our guide to opting out of AI training explains how to do the same at ChatGPT, Gemini and Meta.