On Aug. 25, Anthropic told employees at its San Francisco headquarters to work from home after receiving notice that its security team could strike, an abrupt reminder of how quickly operational risk can intrude on a company sprinting toward an initial public offering. The AI startup had just signed with the largest security firm in North America, and within days it was facing the possibility that the guards guarding its offices would walk off the job.
The dispute is about wages and working conditions, according to people familiar with the matter, and it involves the contracted security staff who control access to Anthropic’s offices, not the company’s own employees. Anthropic said the decision to close the office was precautionary, intended to keep work flowing while the labor issue is resolved. It did not say when the office would reopen.
The episode lands at a delicate moment. Fortune reported the same day that hundreds of Anthropic employees stand to become millionaires when the company goes public, the result of stock grants accumulated over years of working at a startup whose valuation has climbed past $180 billion. The juxtaposition could hardly be starker: a workforce about to collect paper fortunes while the people guarding their desks picket over pay.
Anthropic’s chief executive, Dario Amodei, has publicly wrestled with the wealth effect of the IPO rather than celebrating it. In interviews and internal discussions, he has warned that sudden wealth can distort incentives, pull engineers out of the lab, and turn a mission-driven company into a place people work for the wrong reasons. Fortune reported that Anthropic has gone so far as to test candidates’ tolerance for the possibility that the stock price goes to zero, a screening question meant to filter out applicants motivated mainly by equity upside.
The security strike warning tests that philosophy in a more mundane way. A company that lectures its own staff about the dangers of enrichment is now negotiating with contract workers who want a larger share of the money flowing through the building. Labor experts said the dispute is a sign of the broader pressure building across the tech economy, where the gap between highly paid engineers and the contractors who clean offices, drive buses, and guard lobbies has become a recurring flashpoint.
Anthropic’s rise has been fast even by AI industry standards. Founded in 2021 by siblings Dario and Daniela Amodei after they left OpenAI, the company built Claude, one of the leading AI assistants, and turned it into a business that analysts estimate generates billions in annual revenue. The company has been preparing its IPO for months, hiring bankers, filing paperwork, and answering the kind of questions about business risk that now include the phrase “labor dispute at headquarters.”
The security contract itself was a point of pride. Anthropic said it had hired the largest security firm in North America to protect its offices and its executives, a reflection of the elevated threat environment around AI companies and the value of the intellectual property inside them. The firm’s workforce, many of whom are hourly employees, sees the contract differently: a marquee client with a bottomless budget, and wages that have not kept pace.
The strike notice is the kind of story that rarely surfaces in a prospectus. Companies disclose legal risk, competitive risk, and supply-chain risk, but the day-to-day texture of operations, the security guards, the cafeteria workers, the shuttle drivers, seldom appears in filings. Anthropic’s decision to close its office makes the invisible workforce visible, at least temporarily, to the investors and employees who will watch the company’s public debut.
For the industry, the episode is a preview of governance questions that other AI companies will face as they mature. OpenAI, xAI, and a dozen smaller labs are all scaling toward public markets with similar profiles: enormous valuations, young workforces, and supply chains full of contract labor. The gap between what their founders say about mission and what their contracts pay is becoming a public-relations and labor-relations exposure.
Anthropic declined to comment on the specifics of the security negotiations. People familiar with the situation said the company is working to resolve the dispute quickly, in part because the IPO timeline does not accommodate extended disruptions. Security coverage at the San Francisco office has been maintained at reduced levels, and the company has said it will reassess the work-from-home order as talks progress.
The deeper question the strike warning raises is about the culture Anthropic is building as it scales. Amodei has said repeatedly that he wants Anthropic to remain a company where people work for the mission of making AI safe and beneficial, not for stock options. The test of that principle is unfolding in the lobby, where the people who check badges and patrol the floors are asking for a raise, and the company that measures its employees’ commitment in interview questions is deciding how to answer.
