Anthropic Updates 2026 Usage Policy to Address Deception, Weapons, and Autonomous Actions

Anthropic has published a new version of its Usage Policy effective November 12, 2026, to address evolving model capabilities and observed misuse patterns. The update consolidates rules on deceptive activity, clarifies prohibitions on weapons and surveillance, and introduces new requirements for autonomous physical actions.

Deceptive Campaigns and Elections

A new section titled Do Not Engage in Deceptive Campaigns or Artificial Activity consolidates previous scattered restrictions. It prohibits using Claude to run fake account networks, fabricate news sites, or obscure the source of messages. The elections section, renamed Do Not Undermine Democratic Processes, now specifically targets voter deception and election disruption while removing the blanket ban on personalized vote targeting to allow for legitimate civic work.

Weapons and Surveillance Clarifications

The policy explicitly prohibits using Claude to develop software and components for weapons, including guidance systems for drones. Surveillance rules are tightened to ban tracking individuals without consent and using Claude to determine who to investigate or arrest. These changes reflect existing enforcement practices but provide more precise language regarding prohibited activities.

High-Risk Use Cases and Autonomous Actions

Requirements for high-risk use cases in health, finance, and legal domains remain, mandating a qualified human in the loop and disclosure of AI usage. Following the launch of the Model Hardware Standard, new controls apply when Claude is connected to hardware capable of autonomous physical actions. Operators must be able to observe and stop equipment, which must also maintain a safe state if disconnected.

Abusive Behavior and Regional Restrictions

A new prohibition targets sustained, cruel behavior toward models without discernible purpose, aligning with the existing ability of Claude to end conversations with persistently abusive users. The Supported Regions Policy is clarified to prohibit access for entities incorporated, headquartered, or majority-owned by persons in unsupported regions.