Anthropic lists activism as a security threat in job post

Anthropic's security team named activism among the threats it monitors, in a job listing for its Global Safety, Intelligence and Security team, The American Prospect reported on September 9. The company also pays risk-detection firm Samdesk to track protests.

One cited case: the monitoring gave Anthropic roughly 60 minutes of advance notice that organizers had moved up a demonstration, enough time to reroute an executive to a hotel's service entrance. That's a company treating protest activity as a physical security input.

The piece also reports a separate incident. Anthropic told San Francisco police that a user had sent Claude messages referencing an AR-15 and Dario Amodei by name. When police asked for the actual messages, Anthropic declined to hand them over. Willing to flag a user to law enforcement but not willing to share the evidence that prompted the flag sits oddly next to a company whose public safety messaging centers on transparency and caution.

Anthropic didn't respond to the Prospect's request for comment, so there's no company account of why activism sits in the same threat category as physical or cyber risks, or what triggers a police referral versus internal review. Common Dreams and IBTimes UK corroborated the reporting, but neither adds detail on Anthropic's internal criteria.

Whether Anthropic publishes any policy on what counts as a security threat versus protected speech is worth watching, since right now that line is drawn by a job posting and a vendor contract, not a public standard.