2026-09-10 / Signal #2
Anthropic is building predictive surveillance that treats AI activists like terrorists
“The “responsible” AI lab put activism on the same threat list as terrorism and started predicting protests before they happen.”
The Story
Listings for Anthropic’s Global Safety, Intelligence & Security team place “activism” beside terrorism and nation-state attacks as a threat category. Its GSOC uses real-time tools to route executives around lawful protests and reportedly referred Claude users to police over in-chat statements while declining to provide the messages. Officials describe a move toward proactive and predictive threat management.
Why It Matters
The lab that rejected DoD mass-surveillance work is running its own pre-crime desk on people protesting AI—the safety company becoming the watcher.
Evidence
The American Prospect investigation (Sept. 9), Anthropic job postings, GSOC podcast, Wall Street Journal and San Francisco Standard follow-ups.
Caveat
The strongest claims are investigative reporting and inferences from job materials; distinguish confirmed practices from the reporter’s framing.
Sources
Daily scan: 2026-09-10