Anthropic will ban repeated abuse of Claude under new usage rules

Anthropic will prohibit users from repeatedly abusing Claude under a revised usage policy taking effect on November 12.

The company’s policy update bars “sustained and needless abusive or cruel behavior” toward its models. Anthropic says the rule targets extreme cases in which users repeatedly behave cruelly without a discernible purpose. Ordinary frustration, criticism, dark creative themes, model testing and research remain allowed.

Claude’s ability to end abusive conversations on Claude.ai and Claude Code will remain the main enforcement mechanism. Anthropic has not announced a new automatic account-banning system specifically for this behavior.

The company introduced conversation-ending capabilities for Claude Opus 4 and 4.1 in August 2025 as part of its research into potential AI welfare. It described the feature as a last resort after repeated attempts to redirect an interaction had failed. Ending a chat prevents further messages in that conversation, but users can still start another or continue other chats on their account.

Anthropic said at the time that it remained highly uncertain about whether Claude or other language models could have a moral status deserving protection. Allowing them to leave potentially distressing interactions was an experiment within that research.

The broader usage policy also clarifies restrictions on weapons and surveillance. It explicitly covers weapons guidance and control software, prohibits tracking people without consent, and bars using Claude to recommend who should be investigated, arrested or charged.

New requirements apply when Claude controls equipment capable of causing injury. A qualified operator must be able to monitor and stop it, and the equipment must stop or remain in a safe state if the connection to Claude is lost. Anthropic says most of the wider changes clarify existing rules as its models take on longer, more independent tasks.