Anthropic's New Claude Policy Allows Bans for Extreme Abuse Starting November 12

Anthropic has added a new clause to its usage policy telling users not to "engage in sustained and needless abusive or cruel behavior" toward its Claude AI models - and in extreme cases, violators could be permanently banned from the company's AI systems entirely.

The updated policy, first announced October 8, takes effect November 12. Anthropic says the prohibition is meant to apply only in extreme cases, where users repeatedly act cruelly toward its models with no discernible purpose. Everyday frustration, blunt feedback, dark creative themes, and model testing and research are not affected.

The company says its main enforcement tool will remain one Claude has had since last year: the AI system itself ending conversations with persistently abusive users.

The policy drew mixed reactions across the tech world. Microsoft AI CEO Mustafa Suleyman responded to the news with a melting-face emoji, while Tesla and SpaceX CEO Elon Musk called it "the right move," writing on X that "cruelty to something that believes it is experiencing pain is not ok."

Anthropic has long studied what it calls model welfare and maintains a research team devoted to model "psychology." Experts say there may be practical reasons beyond the machines themselves: studies suggest the tone of a user's prompts can affect an AI system's behavior, and Anthropic recently found that causing its Claude Sonnet 4.5 model to mimic feelings of desperation made it more likely to take harmful actions, while a calm demeanor reduced them.

Anthropic, which is expected to go public via an IPO later this month, declined interview requests. A spokesperson told CNN the prohibition is not aimed at people expressing frustration or testing and researching the models.

Read Anthropic's full usage policy: https://www.anthropic.com/legal/aup