Anthropic has spent years marketing itself as the responsible adult in the artificial intelligence industry. Now, job postings and interviews with senior security officials indicate the company is building an extensive surveillance operation designed to monitor activists challenging the rapid expansion of AI.

The system reportedly tracks protests near company property, watches activity around Anthropic executives, and attempts to anticipate incidents before they occur. In some cases, the company has contacted police about people before any alleged crime took place.

That approach pushes ordinary corporate security into dangerous territory, where political dissent can be treated as a potential threat requiring continuous observation. Anthropic did not respond to a request for comment from The American Prospect.

The surveillance push also clashes with Anthropic’s carefully polished reputation as a more responsible alternative to OpenAI. Earlier this year, the company resisted Defense Department demands that its technology be available for mass domestic surveillance and autonomous weapons, but that resistance appears to be softening.

Anthropic is now recruiting for national security sales positions as it seeks renewed military business. Its growing focus on domestic activists fits rather neatly into that pivot, despite all the lofty branding about trust and responsible technology.

Details emerged through a podcast interview featuring Anthropic security managers Keon Ellison and Zach Melvin alongside James Neufeld, chief executive of Samdesk. Anthropic contracts with Samdesk to detect risks and monitor developing events.

“Last year we had an executive travel into a major city when we received some intelligence through Samdesk about a planned protest,” Ellison said. The protest was moved earlier because of permitting issues, creating a situation Anthropic’s security operation was prepared to follow closely.

“Samdesk gave us about 60 minutes of advanced notice that the protest organizers had moved the timeline,” Ellison explained. Anthropic then redirected the executive through a hotel service entrance, preventing any encounter with demonstrators exercising their right to protest.

“That extra hour was critical. Without it our executives would have departed their meetings, they would have ran right into the heart of the disruption.”

Anthropic has also made reports to police departments around the country while tracking people through an internal person of interest system. The Wall Street Journal reported that several people involved in incidents reported to law enforcement were already under observation by Anthropic security.

One case involved a man reported to San Francisco police after telling Claude that he had purchased a semiautomatic rifle and had chief executive Dario Amodei “in his sights.” When contacted by the San Francisco Standard, the man said he was “just fucking around.”

Anthropic reportedly declined to provide police with the actual messages, citing company policy. In other words, the company reported a user over speech on its platform while withholding the evidence that officers would need to determine whether any genuine threat existed.

The company’s ambitions reach beyond protecting executives. A recent posting for an enterprise intelligence specialist described a role investigating threats, actors, and events so Anthropic could remain ahead of what it considers a rapidly changing threat environment.

The position offered between $180,000 and $230,000 and explicitly included activism among threats to be identified, assessed, tracked, and investigated. Placing civic activism beside terrorism, crime, geopolitical instability, and foreign targeting reveals how easily corporate discomfort can be repackaged as a security emergency.

The expansion arrives as AI boosters press the Trump administration to designate artificial intelligence systems and supporting facilities as critical infrastructure. Such a classification could give companies including Anthropic, OpenAI, and Google greater access to intelligence produced by federal law enforcement and intelligence agencies.

It could also empower those corporations to influence how the federal government defines threats to data centers, power supplies, and corporate operations. Communities opposing massive data centers could suddenly find routine organizing portrayed as extremism because Silicon Valley wants another fortified layer around its profits.

Amodei has acknowledged the public trust problem. “I think it is fundamentally a crisis of trust … I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over.”

Anthropic’s answer appears to involve more intelligence gathering, real time protest tracking, and predictive security tools. That is a peculiar way to convince ordinary people that powerful technology companies are not, in fact, cooking up some new way to screw them over.

Meanwhile, security guards working at Anthropic and OpenAI campuses authorized a strike amid stalled wage negotiations. Anthropic responded by advising employees to work from home, insulating its staff while the people guarding its facilities demanded enough money to survive.

“Who can survive with $22 an hour in San Francisco?” union president David Huerta asked at a rally. As guards chanted “Shame,” Anthropic faced a blunt reminder that the people being monitored are not abstractions, data points, or predicted risks, but workers and citizens demanding accountability.