Anthropic Is Building a Predictive Surveillance System to Monitor Activists
Points and comments are a snapshot, not live.
Anthropic builds predictive surveillance to monitor activists, contradicting its responsible AI image.
Job postings and interviews reveal Anthropic is building a monitoring system to track activists opposing AI development. The system includes real-time protest tracking via Samdesk, a "pre-crime" approach predicting incidents, and reporting suspects to police. Anthropic reported a Claude user to SFPD for threatening CEO Dario Amodei, but refused to show police the messages. The firm is hiring an enterprise intelligence specialist for $180,000-$230,000 to track threats including activism. This contradicts Anthropic's earlier refusal of DoD contracts for mass surveillance, as it now pursues national security sales and critical infrastructure designation for AI.
The article notes Anthropic's own security guards authorized a strike over pay ($22/hr in San Francisco), to which Anthropic responded by telling employees to work from home.
What commenters are saying
Top commenters see Anthropic's surveillance as predictable, noting parallels to Palantir and government contracts. One cites Palantir's Maven system using Claude for targeting in Iran. Commenters split: some argue this reveals closed AI as inherently authoritarian, while others note open models face the same risk. A commenter points to the article's detail that Anthropic only rerouted an executive around a protest, not proactive surveillance. Several express disillusionment with Anthropic's "responsible AI" branding. Lower-ranked comments raise conspiracy theories about CEO Dario Amodei's wife and deep state connections, though other commenters push back on this as unproductive paranoia.