Past year's threat intel reshaped policy: influence ops consolidated, weapons control explicitly named, surveillance catches retroactive tracking too

Summary

Anthropic released its 2026 Usage Policy update, effective November 12, consolidating and clarifying rules against misuse observed over the past year. Key changes include a new unified section prohibiting deceptive campaigns (political or commercial) using fake accounts and fabricated media, narrowed election restrictions focused explicitly on voter deception and election disruption, explicit prohibition of weapons control software and autonomous vehicle arming, and refined surveillance bans covering real-time and retroactive tracking. The policy removes blanket personalized campaign targeting restrictions while maintaining prohibitions on deceptive targeting and data misuse.

Key Points

• New policy effective Nov 12: consolidated deceptive campaigns section
• Explicit ban on weapons control software, drone/autonomous vehicle arming
• Prohibits tracking without consent (real-time and retroactive analysis)
• Removes blanket political targeting ban; allows legitimate civic multilingual work
• Updates driven by threat intelligence: influence ops, weapons dev, surveillance patterns