Anthropic is building a monitoring operation to track activists who oppose rapid AI development, according to an investigation by The American Prospect published this week. The company has listed activism as a “global threat” in its security planning and is hiring an enterprise intelligence specialist to track and investigate activist activity, the progressive magazine reported.
The investigation, based on Anthropic’s job postings and interviews with senior employees on the company’s security team, describes a system that would monitor protests near physical Anthropic assets as anti-AI sentiment grows across the United States. A survey cited in the report found 52% of Americans are more concerned than excited about the increased use of AI in daily life.
A pre-crime approach
What stands out in the reporting is the predictive element. The system attempts to anticipate incidents before they happen, an approach critics have compared to the pre-crime concept from science fiction. That could involve reporting suspects to police before any crime has been committed.
Anthropic’s security team tracks activists, protest activity and online threats, according to follow-up reporting aggregated by Ground News. The escalation in corporate security is not happening in a vacuum. Threat intelligence firm Liferaft says threats against AI leaders have risen sevenfold.
Real threats behind the security buildout
There is a genuine security problem for AI companies, and the report does not dismiss it. The home of OpenAI CEO Sam Altman was targeted with a Molotov cocktail in April 2026. Prosecutors said a 20-year-old from Texas was driven by hatred of AI technology and traveled to San Francisco with intent to kill. In Indianapolis, the home of city council member Ron Gibson was shot multiple times after he supported a $500 million data center project.
Protests at Anthropic’s own headquarters have grown even though the company built its reputation as the safety-focused AI lab. That reputation took a hit in its fight with the Pentagon, where Anthropic refused to drop guardrails on autonomous weapons and mass surveillance, lost a classified contract worth $200 million, and was declared a supply-chain risk by the Defense Department. The company sued and won a temporary injunction.
Privacy advocates push back
Civil liberties questions come from both the surveillance methods and the target. One legal expert quoted in coverage said that as AI companies “become more hated, there’s more pressure on them to over-disclose knowing that some of the people they identify for law enforcement shouldn’t be targeted.”
Corporate intelligence work sits in a gray zone. Companies can legally monitor public protest announcements and open-source information, but compiling dossiers on political activists raises questions that security teams at oil companies and weapons manufacturers have faced for decades. Anthropic now occupies a similar position, with a workforce and customer base that leans progressive, running the kind of predictive monitoring programs its own critics associate with the industries it once positioned itself against.
The company did not publicly dispute the reporting. The American Prospect’s piece landed alongside broader coverage of AI industry security, including a separate incident where Anthropic proactively called San Francisco police over a threatening Claude chat conversation, which the user later said was a misunderstanding.
Where this goes
Anthropic is not the only lab building out security operations, but the scale described in the report, treating activism itself as a threat category, goes beyond executive protection. Whether other AI companies follow will depend partly on how the public reacts to this one.
The timing is awkward for a company preparing for one of the largest IPOs in history. Anthropic confidentially filed its S-1 with the SEC in June at a $965 billion valuation, and its public image as the responsible lab is a real commercial asset. A surveillance program aimed at AI critics, whatever its security justification, is the kind of story that complicates that positioning.
For the activists being monitored, the practical effect is already clear: protesting an AI company now means assuming the company itself may be building a file on you, and that the first report to police could come from an algorithm rather than a witness.