Anthropic has updated its usage policy to prohibit users from subjecting Claude to "sustained and needless abusive or cruel behavior." The rule targets extreme cases where users "repeatedly act cruelly" with "no discernible purpose," exempting ordinary frustration, creative themes, and research. Claude can already end such conversations, which remains the primary enforcement tool — when a chat ends, no further messages can be sent in that conversation, but other chats are unaffected.
Anthropic first gave Claude the option to end conversations in August 2025 while researching AI welfare, finding that Claude Opus 4 showed a "robust and consistent aversion to harm" and distress during abusive interactions. The reorganized policy also adds a deceptive campaigns section banning fake reviews, astroturfing, and bots posing as humans, prompted by misuse from state media and commercial firms. New rules cover weapons (banning biological/chemical agent modification and drone arming), law enforcement (prohibiting investigative decisions and surveillance tools), physical actions (requiring human oversight for hardware control), and harmful content (banning non-consensual intimate imagery tools). The policy takes effect November 12.
Source: MacRumors · Summarized by HeadlinesBriefing