Anthropic builds predictive surveillance operation to monitor activists
New CapabilitiesAI lab tracks protests in real time and reports chatbot users to police over in-app speech
Yesterday: Prospect publishes surveillance investigationNew here? Follow stories to track developments over time. Create a free account to get updates when stories you care about change.
Overview
Updated 1 hour agoAnthropic, the AI lab that positions itself as the safety-first alternative to OpenAI, runs a 24/7 security operation that tracks protest activity near its executives and keeps files on individuals it calls 'persons of interest.' It reports chatbot users to police based on what they type to Claude, and its own security manager describes the goal as predictive, preventative threat management.
The American Prospect's 9 September investigation, built from job postings, police records, and a podcast interview with two of Anthropic's security chiefs, shows the company pays intelligence analysts up to $230,000 to investigate 'activism' as a threat category, alongside terrorism and crime. In one documented case, Anthropic reported a Claude user to San Francisco police for saying he had CEO Dario Amodei 'in his sights,' then declined to show police the actual chat logs.
Why it matters
A private AI company now decides which chatbot conversations become police referrals, with no court review and no public oversight.
Questions about this story
Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.
No questions yet — be the first to ask.
Key Indicators
Voices
Curated perspectives — historical figures and your fellow readers.
Play
Exploring all sides of a story is often best achieved with Play.
Higher or Lower
A number from this story, against one from elsewhere in the news — guess which is bigger, then keep the chain going. 5 rounds, 3 strikes; a miss costs a strike and resets your streak.
Keyboard: ↓/L lower · ↑/H higher
0 points — sign up to put that on the leaderboard.
Connections
Sixteen names from the news. Find the four hidden groups of four. Four mistakes max.
Sign up to keep a daily streak — a new puzzle lands every day.
Exit debate?
Your progress in this debate will be lost.
- 1 Two AI personas square off on this story.
- 2 You predict who'll win each round — correct picks earn XP.
- 3 One crossfire question is yours to fire. Pick it carefully.
Couldn't generate a topic
Select Your Champions
Choose one persona for each side of the debate
DEBATE TOPIC
Choose personas with different perspectives for a more dynamic debate.
Select debater for this side:
No debate personas available right now.
Select debater for this side:
No debate personas available right now.
Who's Got This Round?
Make your prediction before the referee scores
The referee scores both sides on
Round Results
Set the Crossfire
Pick the question both personas must answer in the final round
Debate Oracle! You called every round!
Sharp Instincts! You know your debaters!
The Coin Flip Strategist! Perfectly balanced!
The Contrarian! Bold predictions!
Inverse Genius! Try betting the opposite next time!
XP Breakdown
Prediction History
People Involved
Organizations Involved
AI lab developing the Claude model family; markets itself as a safety-first alternative to OpenAI.
Aggregates open-source data streams to warn corporate clients about protests, disasters, and disruptions in real time.
Washington-based progressive magazine covering politics and policy.
Timeline
2025 September 2026
-
Prospect publishes surveillance investigation
Latest InvestigationThe American Prospect details the full program, including pre-crime approach and police referrals.
-
Job posting seeks intelligence analysts
RecruitingAnthropic posts $180K-$230K role to investigate threats including activism.
-
Anthropic reports chatbot user to SFPD
LegalCompany refers Claude user over alleged threat to CEO; withholds chat logs from police.
-
WSJ reports person-of-interest tracking
DisclosureAnthropic tells the Journal it tracks concerning behavior over time to catch escalation early.
-
Anthropic security chiefs describe protest monitoring
DisclosureIn a podcast, security managers describe using Samdesk for real-time protest tracking and executive rerouting.
Historical Context
3 moments from history that rhyme with this story — and how they unfolded.
COINTELPRO (1956–1971)
The FBI ran a covert counterintelligence program targeting domestic political groups—civil rights organizers, anti-war activists, and others. Agents used informants, wiretaps, and disruption tactics against citizens with no criminal charges.
Exposed in 1971 when activists burglarized an FBI office in Media, Pennsylvania, and leaked documents to the press. The Church Committee investigations followed, leading to restrictions on domestic surveillance.
Generated lasting legal limits on government surveillance of political activity, including the Foreign Intelligence Surveillance Act of 1978.
Anthropic's job posting lists 'activism' as a threat category alongside terrorism and crime. The parallel to state surveillance of political dissent is direct, except this time a private corporation with national security ambitions runs the program.
Elonis v. United States (2015)
Anthony Elonis posted rap lyrics on Facebook describing killing his wife and an FBI agent. Federal prosecutors charged him under a law barring interstate threats, and a jury convicted him.
The Supreme Court reversed the conviction 8-1, ruling that prosecutors must prove the speaker intended the statement as a threat.
Established that online speech can't be punished as a threat without proof of intent, giving First Amendment protection to hyperbolic or joking statements.
Anthropic reported a user to police for telling Claude he had an AR-15 and CEO Dario Amodei 'in his sights.' The user said he was 'just f**king around'—exactly the kind of speech Elonis protects unless intent is shown.
Geofeedia and protest surveillance (2014–2016)
Chicago's Geofeedia sold law enforcement a social media monitoring tool that tracked protesters in real time during the Ferguson uprising and Black Lives Matter demonstrations. The ACLU of California obtained procurement records tying the tool to hundreds of police departments.
After the ACLU exposed the contracts in October 2016, Facebook, Twitter, and Instagram cut off Geofeedia's access to their data streams. The company's core product collapsed.
Showed that private protest-surveillance tools can be dismantled when the public learns how they work and platform access is revoked.
Anthropic's contract with Samdesk mirrors Geofeedia's model—real-time monitoring of protest activity—but the client is a corporate security team, not police. The difference is that Anthropic also holds the conversational data from its own chatbot.
