According to Anthropic, it published an updated Usage Policy on October 8, most of whose changes are meant to clarify existing rules as Claude takes on longer, more independent work. A new section, Do Not Engage in Deceptive Campaigns or Artificial Activity, consolidates restrictions previously scattered across its elections, fraud, privacy and disinformation rules, covering networks of fake accounts, fabricated news sites and the infrastructure behind influence operations. The elections section is renamed Do Not Undermine Democratic Processes, and Anthropic removes a blanket prohibition on personalized vote and campaign targeting while keeping deceptive targeting banned. The weapons section is clarified to include the software and components that make weapons work and actions such as arming drones and other autonomous vehicles. Anthropic also tightens surveillance and law-enforcement language, barring tracking of people without consent and decisions about who to investigate, arrest or charge, and adds controls for high-risk uses in health and finance as well as for autonomous physical actions. The updated policy takes effect November 12.
Our takeOur take: This is mostly housekeeping rather than a change in enforcement, and Anthropic says as much, but the direction is telling. As models take on longer, more autonomous tasks, policy has to shift from banning topics toward governing behavior over time, and consolidating deceptive activity into one place is a reasonable response. The quietly important change is what it loosens: dropping the blanket ban on personalized voter targeting to spare legitimate civic work shows how hard it is to write rules that catch misuse without freezing ordinary use.