The American Prospect reports Anthropic runs internal 'person-of-interest' files on people its security team flags as risks, including activists, and pays a firm for advance protest intelligence, and in August it reported a Claude user to police over a death threat while refusing to hand over the actual messages.
Read the full story: Anthropic Runs a 'Person-of-Interest' System to Track Activists →
Transcript
Anthropic markets itself as the AI lab willing to say no. In February it publicly refused a Pentagon request tied to mass domestic surveillance. But a new report says the company runs its own version internally. Anthropic told the Wall Street Journal it keeps a person of interest file, tracking people over time to catch escalation early. On a podcast, its security team described paying a firm called Samdesk for advance warning before protests.
And a live job posting for an intelligence role lists activism right next to terrorism and nation state threats. In August, Anthropic reported a real death threat against its CEO to police. But the officer's report says the company refused to hand over the actual messages. Safety process, or something else.