Anthropic's first usage policy overhaul in a year bans cruel treatment of Claude and AI surveillance

Anthropic has overhauled the usage policy for its Claude models for the first time in more than a year. The update bans "sustained and needless abusive or cruel behavior" toward Claude and consolidates new restrictions covering election interference, weapons development, surveillance, and health and financial use cases. According to The Verge, the model welfare clause is the most significant change in the revision.

Cruelty toward Claude is now banned

Anthropic frames the new abuse rule narrowly. It is meant to apply only in extreme cases, where users repeatedly act cruelly toward the models with no discernible purpose, and does not cover ordinary frustration, pushback, dark creative themes, or model testing and research. According to The Verge, ending the conversation remains the "primary enforcement mechanism," and Anthropic declined to comment on whether stronger measures such as user bans could follow.

The rule builds on a step the company took last August, when it announced that Claude could terminate conversations with persistently harmful or abusive users as part of its research into model welfare. That research program reflects unresolved questions inside Anthropic: Kyle Fish, who leads the company's model welfare research, told The Verge in February that questions about internal experience, consciousness, moral status and welfare are serious ones under investigation as models become more capable, and CEO Dario Amodei said that same month that the company does not know whether its models are conscious.

The wider industry disagrees. The Verge notes that a September draft of Microsoft AI's code of conduct, posted by Mustafa Suleyman, initially stated that the idea of model welfare is wrong and that AI should not have rights or legal personhood. The document was later changed to a more measured stance that rejects legal personhood and welfare rights while acknowledging that the science of AI consciousness is far from settled.

Elections, weapons and surveillance

Anthropic has also gathered several previously scattered restrictions into a single ban on deceptive commercial or political campaigns. The policy restricts efforts to hide who is behind a message or to amplify content through fake accounts or posts, and its election section prohibits voter deception and election disruption, including misinformation about candidates or how to vote, impersonation of candidates or election officials, and attempts to suppress turnout.

Weapons development was already barred, but The Verge reports that Anthropic has seen multiple attempts to use Claude to develop guidance for and control software for weapons. The updated policy expands the ban to software and components that make weapons work, as well as actions such as arming drones and other autonomous vehicles.

The surveillance rules were sharpened after Anthropic's most recent threat intelligence report pointed to growing use of Claude-powered surveillance to identify and track political dissidents. Tracking people without consent is now prohibited whether it happens in real time or through analysis of previously collected data. Claude also cannot be used to decide or recommend who to investigate, arrest or charge in law enforcement or criminal justice processes, or to build or improve tools designed for surveillance.

Notably, these rules can be modified for contracts with certain governmental customers if, in Anthropic's judgment, the contractual use restrictions and safeguards adequately mitigate potential harms. Anthropic has previously contracted with the US military.

A new rule for physical AI

The policy also adds a hardware clause: when Anthropic's models are connected to hardware that takes autonomous physical actions and might be capable of causing injury, a "qualified operator" must be able to observe the equipment and stop it if needed. The Verge points out that it is unclear whether that operator must be human. The addition reflects the industry-wide push to give AI systems physical embodiments through robotics, and may offer hints about Anthropic's future partnerships in that field.

Why it matters

The update is a snapshot of how AI misuse is evolving in practice. Anthropic is now writing policy against concrete attempts to use Claude for weapons control software, dissident tracking and election manipulation rather than hypothetical risks. It also shows Anthropic formalizing a stance on model welfare that competitors such as Microsoft AI explicitly reject, which makes enforcement the key question: if conversation termination remains the only mechanism, the cruelty ban is largely symbolic. Meanwhile, the government carve-out means the weapons and surveillance restrictions are not absolute, and the new operator requirement signals that Anthropic expects Claude to eventually drive machines that act in the physical world.