By Hamza Faisal
The Pentagon is done asking nicely. In a video interview with The Information, Emil Michael — the Pentagon's chief technology officer, an under secretary of war, and a vice chair of the White House's new AI task force — said the Trump administration has a full range of options if AI companies refuse to cooperate with the government. And one of the biggest sticks, he said, is export controls.
Michael then offered what he framed as a case study: Anthropic. According to his account, the government placed export controls on Anthropic within two hours of the company refusing to pull back a model the government said had been jailbroken.
Two hours to comply — or else
The incident Michael described dates back to June 12, when the Commerce Department issued an export control directive against Anthropic's Fable 5 and Mythos 5 models. The directive — delivered to Anthropic at 5:21 PM ET citing national security authorities — was triggered by what the government said was a method for bypassing, or "jailbreaking," Fable 5.
Anthropic's review reached a very different conclusion: the demonstrated technique only revealed a small number of previously known, minor vulnerabilities — the kind that other publicly available models could surface without any bypass at all. But it complied anyway, abruptly disabling Fable 5 and Mythos 5 for all customers to ensure compliance, disrupting hundreds of millions of users. Access was only restored on June 30, after the company worked with federal agencies on new protocols, agreed to proactively notify the government of malicious activity, and submitted to safety-institute review of its models.
Michael's claim about the two-hour turnaround is his own account and has not been independently confirmed.
A feud that never really ended
The export-control episode was only one front in a bruising standoff between the Pentagon and Anthropic. Earlier this year, the department pushed Anthropic to strip safety guardrails from Claude for military use. The company refused, citing concerns about mass surveillance and autonomous weapons. Defense Secretary Pete Hegseth responded by designating Anthropic a "supply chain risk" — a label normally reserved for foreign companies — and the department pledged to stop using its tools. Anthropic sued the Trump administration, calling the move "unprecedented and unlawful," and a federal judge overturned the ban in August.
Still, a Pentagon official told the BBC the department "has ceased the use of Anthropic products," and Michael says the department will keep Anthropic barred as long as its software carries a "moral principle" that lets its models exert control over humans. Meanwhile, on May 1, the Department of War signed classified-network AI agreements with eight other tech companies — SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, AWS, and Oracle — to build an "AI-first" fighting force.
Export controls are no longer just about chips
What makes Michael's remarks significant is the weapon being named. Export controls were long aimed at hardware — chips, tools, lithography. The June episode showed they can be pointed at software too, and pointed globally: the order covered Anthropic's models for foreign nationals everywhere, but compliance forced the company to pull the models for everyone, everywhere, within hours.
That precedent lands as governments on both sides of the Pacific reach for trade controls over AI. Beijing has been issuing its own export-control announcements in recent days, and Washington's message from its top tech official is now explicit: cooperate, or the next order could name you.
By Hamza Faisal
