Anthropic, the San Francisco-based maker of the Claude AI models, will begin treating persistent cruelty toward its models as a violation of its rules. The company published the latest version of its Usage Policy on Oct. 8, and the changes take effect on Nov. 12.
The new provision prohibits “sustained and needless abusive or cruel behavior toward our models.” Anthropic said it is meant to apply only in extreme cases where users repeatedly act cruelly “with no discernible purpose,” and does not cover ordinary frustration, pushback, dark creative themes, or model testing and research.
The company tied the rule to a feature it introduced in August 2025 that lets Claude end rare conversations with persistently abusive users on Claude.ai and Claude Code. “Claude’s ability to end these interactions will remain the primary enforcement mechanism,” Anthropic wrote. When it launched that feature, the company described it as part of its exploratory work on model welfare and said it remained highly uncertain about the potential moral status of AI models.
Influence operations and elections
Most of the update clarifies existing rules. Anthropic said it has seen state media outlets, government propaganda offices and commercial firms use Claude to run networks of fake accounts and fabricated news sites, and it consolidated those prohibitions into a new section on deceptive campaigns and artificial activity.
Its elections section, renamed “Do Not Undermine Democratic Processes,” now focuses on deceiving voters or disrupting elections, while a blanket ban on personalized voter and campaign targeting was lifted to allow uses such as translating voter information or sending ballot cure notices.
Weapons, surveillance and robots
The policy now spells out that the weapons ban covers guidance and control software and arming drones or other autonomous vehicles. It bars tracking people without their consent and using Claude to decide or recommend who should be investigated, arrested or charged. New requirements also apply when Claude controls hardware capable of causing injury: a qualified operator must be able to stop it, and it must hold a safe state if Claude disconnects.
