Anthropic published a new Usage Policy for Claude. It consolidates rules on deceptive campaigns into one section. The elections section now focuses on undermining democratic processes. Weapons prohibitions now explicitly cover guidance and control software. Surveillance rules now precisely prohibit tracking without consent. High-risk use cases require a human in the loop and user notification. Autonomous physical actions need a qualified operator and safe-state controls. Abusive behavior toward models is now prohibited in extreme cases. Supported Regions Policy clarifies access restrictions. The policy takes effect on November 12.
What changed
Previous policy had scattered rules on deceptive activity, elections, weapons, and surveillance.
What it unlocks
Clearer guidance on prohibited uses and high-risk requirements for Claude.
- takes effect 12 November 2026
Sources