Anthropic will prohibit sustained abusive treatment of Claude

Anthropic has added a prohibition on sustained and needless abuse of its models to its The rules take effect on November 12, 2026.

The provision appears among restrictions on cruel, abusive and psychologically harmful conduct. Lenta.ru, citing The Verge, reports that it concerns extreme cases of repeated treatment without an apparent purpose, rather than ordinary frustration with Claude's answers.

Criticizing the chatbot, arguing with it, discussing dark subjects and researching or testing a model do not in themselves fall under the prohibition. Anthropic links the change partly to research into possible AI welfare. The company considers questions about modern models' consciousness and moral status unresolved.

The revision also clarifies restrictions on disinformation, political manipulation, weapons development and tracking people without consent. Separate requirements address high-risk fields, including medicine and finance. The policy covers Anthropic apps, its API and products that integrate its models.