Anthropic tightens Claude policy on abuse and misuse

Anthropic said it had revised its usage policy and described the change as a response to new misuse cases. It named election interference, weapons development, surveillance, and health and financial uses among the areas covered.

One of the changes barred repeated cruel or abusive treatment of Claude. Anthropic said it had already announced last August that Claude could end conversations with persistently harmful users as part of its model welfare work, and that ending chats remained the main enforcement step.

The company said the rule was meant for extreme cases in which users kept acting cruelly without a clear purpose. It said the limit did not apply to ordinary frustration, pushback, dark creative themes, or testing and research.

Anthropic also brought together several older limits into a broader ban on deceptive commercial or political campaigns. It said that included hiding who was behind a message, using fake accounts or posts, misleading voters, impersonating candidates or election officials, and trying to suppress turnout.

The company said its weapons rules now covered software and components that make weapons work, including arming drones and other autonomous vehicles. It also said Claude could not be used to track people without consent, decide whom to investigate, arrest or charge, or build surveillance tools, and that some rules could change for certain government customers if safeguards were adequate.