Surveillance States · Investigation · September 2026
The self-styled AI safety company is watching its critics.
Anthropic warns AI could end the world — and is also, per new reporting, building a predictive-security operation that lists 'activism' among the threats it tracks, and reportedly flags people to police before anything happens.
FACT
§2 · Thesis
A company whose brand is caution about power it can't control is reportedly building the pre-emptive surveillance apparatus its own philosophy should make it wariest of.
COI up front: Anthropic makes the model used to produce this work — graded the same as anyone, not softened, not inflated. The report and the job posting are FACT; the monitoring scope PROBABLY TRUE; the pre-crime police example SOME SMOKE.
The number
$180–230Kthe 'threat intelligence' hire
the salary on Anthropic's 'enterprise intelligence specialist' posting — to 'identify, assess, track, and investigate global threats including... activism,' alongside terrorism and nation-state targeting.
The American Prospect
§5 · Graded Claim
A job posting lists 'activism' among the global threats Anthropic's security team tracks.
FACT
Anthropic's Global Safety, Intelligence, and Security (GSIS) team posted for an 'enterprise intelligence specialist' ($180-230k) to 'identify, assess, track, and investigate global threats including geopolitical instability, terrorism, crime, activism, nation-state targeting of the AI sector...' It also contracts the risk-detection firm Samdesk. Grouping 'activism' with terrorism as something to investigate is the fact that started the story.
§5 · Graded Claim
It monitors activists and protests near execs — and even Claude users flagged for concerning statements.
PROBABLY TRUE
Sourced to interviews with Anthropic security officials: monitoring activists/organizers near executives, execs' travel and protests near sites, and reviewing individual Claude users flagged for concerning statements; a security manager described moving 'from reactive... to proactive and predictive and preventative threat engagement.' Strong but single-outlet.
§5 · Graded Claim
The report says Anthropic flags people to police before a crime — on one cited example.
SOME SMOKE
The 'pre-crime' claim: Anthropic reports individuals to police before a crime, citing one instance — a Claude user who mentioned an AR-15 and later said he was 'just fucking around.' Serious and specific, but a single outlet and a single example, and Anthropic did not comment. SOME SMOKE: a documented allegation, not an established pattern.
§7 · Why it matters now
When the people warning about the danger build the watchtower.
The uncomfortable core isn't that a company has security — it's that a private AI firm whose identity is caution about uncontrollable power is reportedly building a predictive, pre-emptive surveillance apparatus pointed partly at its critics and its own users. A Surveillance States story with a private-sector face, rhyming with Flock's protest surveillance. Ordinary corporate security is carried as the counterweight; 'pre-criminalizing the opposition' is posed as the question, not a verdict. And to say it again: Anthropic makes the tools we build with — the disclosure is the point.
▸ Surveillance States →