Dario Amodei, Anthropic’s CEO, has a theory about why people distrust AI companies. “Ordinary people don’t trust companies, governments, or the tech industry,” he said recently, “and always suspect that we are cooking up some new way to screw them over.” He’s right about the distrust. What he left out: Anthropic has reportedly built a predictive security operation that formally classifies those skeptical people — once they organize — alongside terrorists and nation-state actors as threats to be tracked, assessed, and managed. This isn’t a camera upgrade. Anthropic created a Global Safety, Intelligence, and Security organization, runs a 24/7 Global Security Operations Center, and is paying $180,000–$230,000 for analysts with OSINT skills borrowed straight from the intelligence community. This resembles how a surveillance app was once built to target political dissidents.
Protest as Threat: How Anthropic Classifies Dissent
Anthropic’s security team reportedly used real-time AI tools to reroute an executive away from a permitted protest — with 60 minutes to spare.
The mechanics are revealing in their ordinariness. In a Samdesk webinar titled “Inside a Modern GSOC,” Anthropic’s GSOC and Travel Security Manager Keon Ellison described a scenario: an executive traveling to a major city, a planned protest whose timeline shifted due to permitting changes, Samdesk delivering 60 minutes of advance notice. The executive was rerouted through a hotel service entrance to avoid “the heart of the disruption.” A peaceful, lawful demonstration — tracked like a weather system, routed around like a security incident.
Then there’s the job posting. Anthropic’s Enterprise Intelligence Specialist role requires analysts to “identify, assess, track, and investigate global threats including:
- geopolitical instability
- terrorism
- crime
- activism
- nation-state targeting”
Activism. Listed between crime and nation-state operations. Critics and labor advocates argue that categorizing activism alongside terrorism reflects a political rather than a safety calculation — and the $180,000–$230,000 salary signals this is no ordinary corporate security hire.
“Transforming operations from reactive information gathering to proactive and predictive and preventative threat engagement and management.” — Anthropic security program manager, as reported by The American Prospect
“Predictive and preventative” is the tell. It means action before an offense. Protest organizers reportedly get flagged and tracked before breaking a single law — a person-of-interest process Anthropic itself confirmed to The Wall Street Journal, describing how it catches “escalation patterns early” and makes routine police reports nationally. In one documented case, Anthropic contacted San Francisco police about a Claude user who sent a threatening message to the chatbot — then reportedly refused to show officers the actual chat logs, citing internal policy. Such tactics echo how corporations have been caught secretly tracking users without meaningful accountability.
Meanwhile, advocacy group Americans for Responsible Innovation is lobbying to classify AI infrastructure as “critical infrastructure.” That designation could give labs like Anthropic privileged access to government intelligence and potentially reframe civic opposition as a national security threat.
The $22-an-Hour Guard and the $230,000 Analyst
The workers physically securing Anthropic’s campuses are fighting for basic wages while the company builds a sophisticated intelligence operation above their heads.
Security guards patrolling Anthropic and OpenAI campuses have authorized a strike over stalled pay negotiations. SEIU-USWW president David Huerta asked plainly: “Who can survive with $22 an hour in San Francisco?” Anthropic’s reported response was an internal email advising employees to work from home during the labor action — keeping the building quiet while the intelligence apparatus tracked the next potential disruption.
“I think it is fundamentally a crisis of trust.” — Dario Amodei, Anthropic CEO
Amodei diagnosed the disease with uncomfortable precision. The responsible AI company is reportedly hiring national-security analysts to monitor the people who don’t trust it — which means he just didn’t mention that Anthropic was already part of the symptoms.
