Anthropic Usage Policy adds Claude cruelty ban, effective Nov 12

Anthropic has added a ban on "sustained and needless abusive or cruel behavior toward our models" to its Usage Policy, which takes effect November 12, 2026.

The clause is narrow. "It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research," the company writes in its update post, reserving the rule for extreme cases where users repeatedly act cruelly with no discernible purpose.

The rest of the refresh reads as operator-tightening. The elections section is renamed "Do Not Undermine Democratic Processes" and focused on voter deception, candidate and official impersonation, and turnout suppression; the earlier blanket prohibition on personalized campaign targeting is gone. Fake accounts, fabricated news sites, and influence operation infrastructure now sit under one consolidated rule against "deceptive activity of any kind (whether political or commercial)."

The weapons prohibition explicitly reaches "software and components that make weapons work," including arming autonomous vehicles, and new hardware language requires qualified operators able to stop equipment and keep it in a safe state on disconnect. On surveillance, Anthropic says Claude cannot "be used to decide or recommend who to investigate, arrest, or charge" in law enforcement contexts, and bars tracking people without consent whether in real time or through historical data.

High-risk use cases get a longer list and a sharper requirement. Candidate screening, loan pricing and claim decisions are now treated as high-risk, needing a qualified reviewer with authority to change or modify the model's recommendation, and the person on the receiving end has to be clearly told AI was used to produce it. Two of the voices in our Who's Who tracker had already linked the policy by the time we went out.