The Trump administration has a new pressure point for the AI industry, and it was not built for software companies.
A senior Pentagon official said the administration may use export controls against AI companies if security concerns arise. Export controls are the rules that decide what technology can leave the country. They are usually aimed at foreign rivals, not domestic labs.
What the Pentagon actually said
The remarks came from Emil Michael, the Under Secretary of War for Research and Engineering. He spoke in an interview dated October 18, 2026.
Michael said the administration is prepared to enforce such controls as needed. The target would be AI labs that do not align with US security protocols.
The warning was conditional. Michael framed export controls as a possible response to security concerns, not as a policy already in force against any company.
He did, however, make clear that the government is willing to move fast. He said the administration intends to act decisively on real threats.
The Anthropic precedent
Michael did not have to speak in hypotheticals. He pointed to Anthropic as an example of how quickly consequences can arrive.
Earlier in 2026, Anthropic was blacklisted within two hours of refusing to address security issues.
The Pentagon also designated Anthropic a “national security supply-chain risk” in early 2026. That label led to a phase-out of the company’s tools.
When Michael warns that non-cooperative labs could face immediate repercussions, the industry has a recent case showing what “immediate” can mean.
Who the Pentagon is working with instead
In 2026, the Trump administration pursued deals to put AI models from OpenAI, Google, and xAI on classified networks.
Companies that cooperate with Pentagon security requirements get access to some of the government’s most sensitive work. A company that refused found itself labeled a risk and phased out.
The China backdrop
Around the time of Michael’s interview, China proposed new export controls covering advanced technologies, including AI, biotech, and solar.
So both Washington and Beijing are reaching for the same policy tool in the same week. Export controls have become a default instrument for governments trying to protect technological leadership.
What this means for the AI industry
The most direct implication is that the relationship between frontier AI labs and the national security establishment is becoming less voluntary.
Export controls would raise the stakes considerably. A procurement ban limits who a company can sell to inside the government. An export restriction could limit where a company can sell at all.
Michael did not lay out what specific security concerns would trigger controls, or which models or products could be covered. The definition of “cooperation” will matter a great deal to every lab weighing how to respond to government requests.
The administration has now shown it can move against a major AI developer in hours. The open question is whether the next move would come from the procurement office or the export desk.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.