Anthropic Predictive Surveillance and Activist Monitoring

Anthropic is implementing a predictive surveillance system to monitor activists and potential threats

Anthropic is building an extensive monitoring apparatus designed to track activists who oppose the rapid development of artificial intelligence and to predict security incidents before they occur. This "pre-crime" approach involves tracking "persons of interest" to identify escalation patterns and reporting suspects to law enforcement before a crime has been committed.

Proactive Threat Detection and "Pre-Crime" Operations

Anthropic's security strategy has shifted from reactive information gathering to "proactive and predictive and preventative threat engagement and management." According to internal security officials, the goal is to achieve "operational maturity" in protecting high-value targets.

Key components of this system include:

• Person-of-Interest Tracking: Anthropic tracks "concerning behavior over time" to catch escalation patterns early. The company has confirmed that several individuals reported to police were already being tracked by internal security.
• Law Enforcement Integration: The company makes routine reports to police departments nationwide regarding perceived threats.
• Predictive Routing: Using intelligence from third-party contractors like Samdesk, Anthropic monitors planned protests in real-time to reroute executives and avoid physical disruptions.

Intelligence Gathering and Global Security Operations

To support these operations, Anthropic's Global Safety, Intelligence, and Security (GSIS) team is expanding. A recent job posting for an enterprise intelligence specialist (with a salary range of $180,000 to $230,000) explicitly lists the identification, assessment, and tracking of "activism" alongside terrorism, crime, and nation-state targeting as a core responsibility. This role utilizes Open Source Intelligence (OSINT) collection and deep-dive research to keep employees in a "defensible position."

Contradictions in Corporate Positioning

These surveillance efforts appear to conflict with Anthropic's public image as a "responsible" AI lab. While the company previously resisted Department of Defense (DoD) contracts involving mass domestic surveillance and autonomous weapons, it has recently begun hiring for "national security sales" positions to restart military contracts.

Furthermore, the company's approach to transparency with law enforcement is inconsistent. In one instance reported by the San Francisco Standard, Anthropic reported a user to the San Francisco Police Department for threatening CEO Dario Amodei via Claude, but refused to provide the police with the actual messages, citing internal policy.

Critical Infrastructure and National Security

There is a broader movement to designate AI and its supporting infrastructure (data centers and power supplies) as "critical infrastructure." Such a designation would place AI labs on the same footing as water and electricity utilities. This shift would likely increase the access frontier labs have to federal intelligence products and potentially allow these companies to influence how federal agencies define "threats" to their operations.

Community Perspectives and Counterpoints

Public reaction to these revelations has been polarized. Some critics argue that the system is a tool for crushing dissent and an example of the "crisis of trust" described by CEO Dario Amodei.

"Democratic protest is now a high-stress situation?"

Others argue that these practices are standard for any large corporation protecting high-value targets and executives from physical threats.

"If I strip out the editorializing... I think the conclusions are: 1. Someone told Claude they had an AR-15 and the CEO of Anthropic was 'in their sights'. Anthropic reported this to the police. 2. Anthropic contracts with a security agency. 3. Anthropic is looking to build in-house capability for security."

Additional concerns raised by observers include the potential for "linguistic fingerprinting"—using AI to identify individuals across different platforms based on writing style—to enhance the scale of this monitoring.