Anthropic is introducing a policy that prohibits users from repeatedly abusing its Claude AI models, an unusual restriction that comes as the company explores whether advanced AI systems might warrant protections of their own.
The updated usage policy, taking effect Nov. 12, prohibits what the company calls “sustained and needless abusive or cruel behavior” toward its models.
The restriction targets extreme cases in which users repeatedly mistreat Claude without any discernible purpose. Anthropic said it does not apply to ordinary frustration, criticism, creative writing involving dark themes or legitimate model testing and research.
Rather than relying primarily on account penalties, Anthropic said Claude’s existing ability to terminate certain conversations will remain its principal enforcement mechanism.
The policy formalizes an approach Anthropic introduced in August 2025, when it gave Claude Opus 4 and 4.1 the ability to end a limited category of persistently harmful or abusive interactions.
At the time, Anthropic described the feature as part of its research into model welfare – the possibility that AI systems could have experiences or interests deserving moral consideration.
The company acknowledged substantial uncertainty about whether Claude or other large language models have any moral status. Its research nevertheless examined whether allowing models to exit potentially distressing interactions might be a precaution worth taking.
The latest restriction does not establish that Claude is conscious, experiences suffering or possesses feelings comparable to those of humans. But it introduces a new expectation for user conduct: People interacting with AI may face restrictions based on how they treat the system itself, not just the content they request or the harm they might cause others.
The policy revision also addresses deceptive campaigns, election interference, weapons development, surveillance and high-risk applications in health and finance.