2026 Usage Policy update
Anthropic Anthropic ● Covered by 3 sources
Anthropic updated Claude’s usage rules and they take effect on Nov. 12. The big theme: tighter wording for risky uses, deception, weapons, and abuse.
Based on reporting by Anthropic, Anthropic — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has published a new version of its Usage Policy, saying the update mainly clarifies rules that were already there. The company says Claude has taken on longer, more independent work over the past year, so the policy now comes with new examples and sharper language for how those rules apply.
One of the biggest changes is a new section called Do Not Engage in Deceptive Campaigns or Artificial Activity. Anthropic says it had already banned this kind of behavior, but the rules were spread across several parts of the policy. Now they sit together, covering fake accounts, fabricated news sites, hidden identity in messaging, and the tools needed to run influence operations, whether the goal is political or commercial.
The elections section got a similar cleanup. It now focuses on keeping Claude from being used to deceive voters or disrupt elections, such as by spreading false claims about candidates or voting, impersonating election officials, or suppressing turnout. At the same time, Anthropic removed a blanket ban on personalized vote and campaign targeting, saying that rule had swept up legitimate work like multilingual voter information and ballot cure notices.
The weapons language is getting more explicit too. Anthropic says people have tried to use its models to build guidance and control software for weapons, so the policy now spells out that its ban includes the software and components that make weapons work, along with actions like arming drones and other autonomous vehicles. The company says this reflects how it has already been enforcing the rule.
There are also clearer limits around surveillance, criminal justice, and high-risk use cases in health, finance, legal rights, and other sensitive areas. Anthropic says Claude can’t be used to decide or recommend who to investigate, arrest, or charge, and it now adds hardware-related requirements for systems that can take autonomous physical actions and might cause injury. The update also adds a ban on sustained, needless cruelty toward the models, while saying that ordinary frustration, pushback, dark creative work, and testing are not the target. The new policy takes effect on November 12.
My take — AI-written commentary, not fact-checked reporting
This is what mature AI policy looks like: less grandstanding, more red pen. Anthropic is doing the boring but necessary work of turning fuzzy prohibitions into something operators can actually read, which is how safety rules stop being decorative. The bigger story is that frontier AI is now messy enough to need policies for drones, fake news farms, and people being rude to chatbots like they owe them money.
Read more about this at: Anthropic