Updated
Updated · The American Prospect · Sep 9
Anthropic Builds Predictive Surveillance System for AI Activists, Reporting Some Suspects to Police
Updated
Updated · The American Prospect · Sep 9

Anthropic Builds Predictive Surveillance System for AI Activists, Reporting Some Suspects to Police

3 articles · Updated · The American Prospect · Sep 9

Summary

  • Job postings and security interviews show Anthropic is expanding a predictive monitoring program that tracks anti-AI activists, watches protests near executives and facilities, and in some cases reports people to police before any crime occurs.
  • A podcast with Anthropic security managers described using contractor Samdesk to detect a protest schedule change 60 minutes early, reroute an executive, and avoid what the company called a disruption through early intelligence.
  • July reporting by The Wall Street Journal said Anthropic tracks “concerning behavior” through a person-of-interest process and had already been monitoring several people later reported to police.
  • One recent case involved a Claude user Anthropic reported to San Francisco police over AR-15 threats against CEO Dario Amodei, though the company then declined to provide the underlying messages, citing internal policy.
  • A $180,000-$230,000 intelligence job posting ties activism monitoring to broader threats including terrorism and nation-state targeting, underscoring Anthropic’s shift toward a national-security posture even as it brands itself a responsible AI lab.

Insights

Why did Anthropic alert police about an assassination threat but refuse to hand over the AI chat logs that triggered the alarm?
How does an AI company justify using predictive surveillance systems to track peaceful activists alongside actual terrorist threats?