Anthropic Is Building a Predictive Surveillance System to Monitor Activists
Anthropic is developing a predictive surveillance system to monitor activists, including those who oppose AI and data center development, and is reporting individuals to police. The company's security team tracks "persons of interest" and uses AI tools for early detection of protests and threats, raising concerns about privacy and due process.
Anthropic, a prominent AI lab, is reportedly developing an extensive surveillance system to monitor activists, particularly those critical of rapid AI development. Job postings and interviews with security officials indicate the firm is adopting a "pre-crime" approach, attempting to predict and prevent incidents by tracking individuals and reporting potential threats to police departments nationwide. This strategy is highlighted by an incident where Anthropic's security team, utilizing intelligence from contractor Samdesk, rerouted an executive to avoid a planned protest in a major city.
The company's surveillance efforts extend beyond executive protection, involving a Global Safety, Intelligence, and Security (GSIS) team tasked with investigating global threats, including activism, and performing "OSINT collection on specific threats, actors, and events." Anthropic's security program manager explicitly articulated the goal of transforming operations from reactive to proactive and predictive threat management.
One notable incident involved Anthropic reporting a user to San Francisco police for in-platform speech, but then refusing to provide the police with the actual messages, citing internal policy. This raises questions about due process and the evidence basis for such reports. The article also notes a potential shift in Anthropic's stance on military contracts, hiring for "national security sales" positions, which aligns with its renewed focus on threat monitoring.
The expansion of Anthropic's intelligence gathering coincides with discussions around designating AI and its infrastructure, including data centers and power supply, as critical infrastructure. Such a designation, advocated by groups like Americans for Responsible Innovation, could afford AI labs increased access to intelligence from federal agencies and potentially lead to the branding of civic opposition, including anti-data center activism, as extremist or terrorist threats. This initiative is unfolding amidst internal labor disputes, as security guards at Anthropic and OpenAI authorized a strike over pay in the Bay Area.