Anthropic tightens its usage policy

Anthropic has revised its usage policy for the first time in more than a year, covering newer high-risk misuse, including election interference, weapons development, surveillance, and health and financial uses. The Verge reports that one of the main changes prohibits sustained and needless abusive or cruel behavior toward Claude. Last August the company said it would let Claude end conversations with persistently harmful or abusive users, as part of research into model welfare. The update says ending a conversation is still the primary enforcement mechanism. Anthropic did not comment on further steps, such as banning an account. It wrote that the rule is meant only for extreme cases, where users repeatedly act cruelly toward the models with no discernible purpose. It does not apply to ordinary frustration, pushback, dark creative themes, or model testing and research.

Beyond model welfare, Anthropic gathered scattered limits into a new ban on deceptive commercial or political campaigns. It restricts efforts to hide who is behind a message, or to amplify content through fake accounts or posts. The election section prohibits voter deception and election disruption, including misinformation about candidates or how to vote, impersonating candidates or election officials, and trying to suppress turnout. The company says the policy has always publicly banned weapons development with Claude, but it has seen multiple attempts to use Claude for guidance and control software for weapons. The new policy extends the ban to software and components that make weapons work, and to actions such as arming drones and other autonomous vehicles. Its last threat-intelligence report suggested growing use of Claude-powered surveillance to identify and track political dissidents, so the surveillance rules are now more explicit. Tracking people without consent is prohibited, whether in real time or by analyzing data already collected. Claude cannot be used to decide or recommend who to investigate, arrest, or charge in a law-enforcement or criminal process, and it cannot be used to build or improve tools designed for surveillance.

The rules can be modified for contracts with certain government customers. Anthropic wrote that it may do so if, in its judgment, the contractual limits and safeguards are enough to reduce the potential harms. The company has previously contracted with the US military. A further rule says that when a model is connected to hardware that takes autonomous physical actions and might cause injury, a qualified operator must be able to watch the equipment and stop it if needed. The Verge notes that the article does not say whether that operator must be a person.