Anthropic Is Building a Predictive Surveillance System to Monitor Activists

gist

Daniel Boguslaw reports that Anthropic is building predictive security to track activists near executives and facilities, using contractors like Samdesk for real-time protest intel and a person-of-interest process that feeds police reports. Job postings for its GSIS intelligence role list activism alongside terrorism and nation-state threats. The piece contrasts that apparatus with Anthropic's public resistance to military mass surveillance, and links the expansion to critical-infrastructure lobbying that could reframe civic AI opposition as a security problem.

ideas

• Predictive security is the stated goal. Samdesk advance notice reroutes executives around protests, while Anthropic tracks escalation patterns and reports people to police before crimes occur.
• Activism is an explicit intelligence target. A GSIS enterprise-intelligence posting asks hires to investigate global threats including activism, OSINT, and nation-state targeting of the AI sector.
• The responsible-lab brand is under strain. After a DoD fight over mass domestic surveillance and autonomous weapons, Anthropic is hiring for national-security sales while expanding domestic dissent monitoring.
• Critical-infrastructure framing raises the stakes. If AI labs become hardened national-security assets, civic opposition to data centers or AI itself can be cast as extremism.
• Platform speech becomes a police tip without full evidence. Anthropic reported a Claude user over AR-15 and Amodei threats, then refused to show police the messages under internal policy.

quotes