Anthropic usage policy update bans abuse toward Claude

Anthropic on Thursday, October 8, published an update to its Usage Policy. It is the company's first revision in more than a year, and it adds a rule against "sustained and needless abusive or cruel behavior" toward its Claude models. The update also gathers existing rules against deceptive campaigns into one section and spells out more clearly what counts as banned weapons and surveillance work. It takes effect November 12, 2026.unite+2

The company said most of the changes clarify rules it already had. It said the revision reflects a year in which Claude took on longer and more independent work, along with customer feedback and misuse patterns described in its September threat intelligence report.anthropic+1

A rule protecting the model

Anthropic said the abuse rule "is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose." It said the rule "does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research". The main way the rule will be enforced is a feature launched in August 2025: Claude can end conversations with persistently abusive users. Anthropic built that feature as part of its research into potential "model welfare". According to The Verge, the company did not say whether other penalties, such as account bans, could follow.theverge+1

The change comes as Anthropic keeps exploring whether its models might have some kind of inner experience. CEO Dario Amodei said on a podcast in February, "We don't know if the models are conscious". Other companies disagree. A draft code of conduct from Microsoft AI, posted by Mustafa Suleyman in September, said, "The idea of model welfare is wrong".theverge

Deception, weapons and surveillance

A new section titled "Do Not Engage in Deceptive Campaigns or Artificial Activity" brings together rules that were previously spread across the policy. It covers hiding who is behind a message, boosting content through fake accounts, and building tools for influence operations. Anthropic said state media outlets, government propaganda offices and commercial firms have used Claude to run these kinds of networks. The company also dropped its blanket ban on personalized vote targeting, saying the rule had blocked legitimate civic work such as translating voter information.unite

The weapons section now explicitly covers "software and components that make weapons work," including arming drones and other autonomous vehicles. Anthropic said it had seen attempts to use Claude to build guidance and control software for weapons. The policy bans tracking people without their consent. It also bars using Claude to decide or recommend who to investigate, arrest or charge. Contracts with some government customers can still change these restrictions if Anthropic judges the safeguards to be adequate, The Verge reported.theverge

Hardware controls

A new requirement covers models connected to hardware that can take physical actions on its own and might cause injury. A "qualified operator must be able to observe the equipment and stop it if needed," and the equipment must be able to hold a safe state if Claude is disconnected.unite+1