2026 Usage Policy Update

Anthropic’s 2026 Usage Policy update reorganizes and clarifies rules to match Claude’s growing capabilities and new misuse patterns. Key structural changes include a consolidated prohibition called Do Not Engage in Deceptive Campaigns or Artificial Activity that brings together prior bans on disinformation, fraud, and covert influence operations. Election-related restrictions are tightened and renamed Do Not Undermine Democratic Processes, explicitly forbidding voter deception, impersonation of officials, and turnout suppression while removing a blanket ban on personalized targeting to allow legitimate civic uses. Weapon-related prohibitions now explicitly cover software, control systems, and actions like arming drones or autonomous vehicles. Surveillance and criminal-justice language is more precise: non-consensual tracking (real-time or retrospective), use of Claude to recommend investigations or charges, and building surveillance-improvement tools are prohibited, while consented monitoring, moderation, journalism, and legal research remain allowed.

High-risk use cases retain requirements for a qualified human-in-the-loop and disclosure that AI was used; the policy now lists which recommendations trigger those requirements and adds hardware safeguards when models control equipment that can cause injury (observable operators, emergency stop, safe-state on disconnect). A new ban addresses sustained, needless abusive behavior toward models and permits Claude to end persistently abusive interactions. The Supported Regions policy is clarified to bar access by people or entities physically located in, incorporated in, or majority-owned by actors in unsupported regions. The update takes effect November 12 and emphasizes continued iterative revisions.