Anthropic tells San Francisco staff to work from home over a strike the union says was never called

Anthropic cleared its San Francisco offices for two days after Allied Universal, the contractor that staffs the buildings' security, warned that a strike by its guards might be coming. The Service Employees International Union, which represents those security officers, said it never called one.

The stay-at-home order covered Monday and Tuesday, according to Business Insider, which cited a person familiar with the matter. The notice from Allied Universal flagged a possible job action at Anthropic's San Francisco buildings. Anthropic declined to comment on the order. Allied Universal did not respond to requests for comment, Business Insider said. The story has circulated in operator circles on Hacker News, where the discussion has focused less on the labor dispute than on what a $1.5 trillion lab treats as enough warning to send its staff home.

SEIU, the union that represents Allied Universal security officers in California, pushed back on the threat characterization. The union told Business Insider it had not called a strike authorization vote and had not issued any strike threats for the week in question. That denial sits in tension with the contractor's reported warning and with Anthropic's decision to shutter its own offices anyway.

The two sides are, however, in the middle of a real labor fight. SEIU has been bargaining with Allied Universal and other California security firms since April, according to the union's security division. Security officers have pressed for higher wages, better healthcare, and more comprehensive job training, SEIU said. The current dispute is not over a new walkout, per SEIU, but the underlying contract talks cover the security officers covered by the Allied Universal agreement.

Anthropic's reaction treats the warning as serious enough to disrupt work, even without a confirmed job action. The company runs a hybrid in-office policy that asks staff to be in the office at least 25% of the time, according to its job postings as reported by Business Insider. Asking engineers, researchers, and operations staff to work from home for two days at a company of Anthropic's size and profile carries a real cost in coordination, and a real signal about how the lab reads its own perimeter.

That signal is larger than the incident itself. Anthropic and OpenAI have faced several threats against their employees in recent years, the Wall Street Journal has previously reported, and security spending has climbed across the tech industry. A preemptive WFH order at a frontier-AI lab, on the basis of a contractor's notice the union disputes, is the kind of move that treats the security perimeter as load-bearing infrastructure rather than a back-office function.

The trigger for the order remains the disputed strike notice, not those broader threats. Allied Universal's communication to Anthropic is the only direct on-record account of an impending action; the contractor has not spoken publicly. The NLRB docket on case 20-CB-373545 is a National Labor Relations Board case file covering SEIU's labor-process filings against the contractor, not direct evidence of an active job action at Anthropic.

The timing puts the company in a delicate position. Anthropic filed a confidential S-1 with the Securities and Exchange Commission in June and is expected to file publicly for its initial public offering as soon as August. On secondary markets, investors have priced the company at $1.5 trillion. A two-day WFH order against a possible strike the union says was never called is not, on its own, a material event. The way the company handled it is.

Allied Universal has not responded to requests for comment. The next concrete data point in the picture is the company's public S-1 filing, which could land any week.