Anthropic Policy Prohibits Sustained, Needless Abuse of Its Models

Anthropic’s Usage Policy prohibits sustained and needless abusive or cruel behavior toward its models, while also restricting election deception, weapons software, surveillance and certain high-risk physical actions.

The rules apply across Claude.ai, Claude Code, Anthropic’s API and developer platforms, cloud-provider access, and products that integrate Claude. Anthropic may warn users, throttle or limit access, suspend accounts or terminate access when it suspects a policy violation.

The policy bars deceptive campaigns involving fake personas, undisclosed sponsorship or coordinated inauthentic behavior. It also prohibits false information about candidates or voting procedures, impersonation of candidates or election officials, and efforts to suppress turnout.

Under the weapons provisions, users may not develop or operate software for targeting, fire-control or engagement systems. The rules also prohibit integrating weapons with drones or other autonomous platforms.

The policy restricts tracking people without their consent through physical location data, biometrics, online activity, communications or previously collected information. For high-risk physical actions, a qualified person must observe the equipment, retain the ability to stop it at any time and maintain independent safety limits.

The policy page lists Nov. 12, 2026, as its effective date. Anthropic previously gave some Claude models the ability to end conversations with abusive users as part of its research into model welfare.