Anthropic requires an operator stop for hardware Claude controls

Anthropic rewrote its Usage Policy to put every fake-account and influence-operation rule in one section and to require that a qualified human be able to stop any equipment Claude controls. The new rules apply when Claude controls hardware that can take autonomous physical actions.

Anthropic published the new version on Oct. 8, 2026, and it takes effect Nov. 12. Anthropic said the weapons and surveillance edits reflect existing enforcement. It said it will keep revising the policy as capabilities and risks change.

One Section Now Covers Deceptive Campaigns

Anthropic moved its rules on fake accounts, fabricated news sites and influence operations out of the elections, fraud, privacy and disinformation sections into “Do Not Engage in Deceptive Campaigns or Artificial Activity.” It covers building tools and infrastructure for such campaigns, and the post says it “applies to deceptive activity of any kind (whether political or commercial).”

The elections section is renamed “Do Not Undermine Democratic Processes” and focuses on deceiving voters or disrupting elections. Anthropic removed the blanket ban on personalized vote and campaign targeting. Deceptive targeting and misuse of voters’ personal data stay prohibited under other sections.

Surveillance and Weapons Rules Name More Cases

The surveillance and law enforcement rules now bar tracking people without consent, whether in real time or through analysis of previously collected data. Claude may not decide or recommend who to investigate, arrest or charge, and may not be used to build or improve surveillance tools. Consented tracking such as fraud monitoring, content moderation, journalism and legal research is permitted.

The weapons prohibition now explicitly covers software and components that make weapons function, and arming drones or other autonomous vehicles.

Claude-Controlled Hardware Needs a Human Stop

New requirements apply when Claude controls hardware that can take autonomous physical actions. “A qualified operator must be able to observe the equipment and stop it if needed,” the policy says. The equipment must also hold a safe state if Claude is disconnected.

Human-in-the-loop and disclosure requirements for high-risk use cases are unchanged. That section now lists which recommendations are covered and which are not.

Extreme Cruelty Toward Claude Is Banned

Anthropic now prohibits sustained, needless cruelty toward Claude with no discernible purpose. The rule targets only extreme cases and excludes ordinary frustration, pushback, dark creative themes, and model testing or research. “Claude’s ability to end these interactions will remain the primary enforcement mechanism,” the post says.

Supported Regions Rules Are Restated

Anthropic restated the prior year’s Supported Regions restriction with clearer enforcement details on the Supported Regions page. It bars use by people physically located in unsupported regions, by entities incorporated or headquartered there, and by entities majority-owned or controlled by persons or entities in those regions.

Our View

We think the safe-state rule puts a design job on the developer. The policy says equipment must hold a safe state if Claude is disconnected. A team building a robot arm on Claude therefore has to decide what the machine does when the connection drops. Claude cannot make that decision once it is cut off.

The operator rule adds a second job. A qualified operator must be able to observe the equipment and stop it. That person needs a way to see the machine and halt it. Teams planning to run such equipment after Nov. 12 should settle both.

The rule protects anyone only if the safe state is something the equipment does on its own when the link to Claude is lost.