The update adds a prohibition on sustained and needless abusive or cruel behavior toward its models, limited to extreme cases of repeated cruelty without discernible purpose. Anthropic says ordinary user frustration, pushback, dark creative themes, model testing, and research are outside that prohibition. The company says Claude’s ability to end persistently abusive conversations will remain the primary enforcement mechanism.
The policy update also consolidates restrictions on deceptive campaigns and artificial activity, refines election-related rules, clarifies weapons and surveillance prohibitions, and specifies requirements for high-risk uses, including models connected to autonomous physical equipment. Anthropic characterizes several of these provisions as clarifications of existing rules and enforcement practice.