Pentagon’s Emil Michael Vows Export Controls if AI Labs Don’t Cooperate With Government

Emil Michael, a senior Department of Defense official and member of President Trump’s “super intelligence” task force, vowed in an interview to use a range of executive tools including export controls to check the behavior of AI companies if they don’t cooperate with the government, while comparing frontier labs calling for more regulation to drunk drivers who demand someone else pull them over.

Michael drew widespread attention as the Trump administration official leading a public feud with Anthropic, pushing for the Defense Department to label the AI company a supply chain risk after it refused to remove safety guardrails. Now, as one of the four senior administration officials leading the Super Intelligence Force that Trump announced this month, Michael is charged with serving as the government’s liaison to Anthropic and other AI companies to develop protocols for handling cyber breaches, hacks and other critical incidents.

And as the Under Secretary of Defense for Research and Engineering, Michael is working to modernize the hulking Defense Department through exploring loans to key supply chain sectors like drone parts and microelectronics.

In a wide-ranging interview with The Information on Friday, the day after the task force’s first official meeting, Michael pushed against the “doom and gloom” mentality pervading public sentiment around AI while arguing for the Trump administration’s responsibility to rein in companies like Anthropic.

Much of the administration’s approach to AI oversight has focused on voluntary mechanics, like the non-binding joint agreement signed between top labs after a White House summit in late September. Still, Michael argued that more stringent checks on AI companies are appropriate if needed, such as the export controls imposed on Anthropic in June that cut off access to its models to non-U.S. users.

“We have lots of tools, whether it’s a restriction on a type of technology being exported because of the advantage we think it might give an adversary to straight up export controls and limitation on foreign nationals’ ability to use that software,” Michael said. “You can count on this administration, this president, to act with urgency should there ever be a real issue.”

Michael brushed off the idea that the Defense Department’s ongoing legal battle with Anthropic would affect his ability to work with the company. He said that both sides are “adults” and that he’s seen Anthropic co-founder Tom Brown, who has recently acted as the company’s main point person in D.C., “many, many times.”

“My first duty is to the Department and to the American people from a national security standpoint,” Michael said. “And if that means I’m talking to someone who’s suing me, well then I’m talking to them.”

Criticism of Anthropomorphizing AI Models

That doesn’t mean Michael has changed his mind on Anthropic. He argued against the idea of anthropomorphizing AI models, such as Anthropic’s policy decision to allow its chatbot Claude to end conversations with abusive users. “If you don’t want humans to lose control you can’t give models the control over humans,” Michael said, arguing that he would not allow Anthropic within the Defense Department “as long as that moral principle is embedded in their software.”

Michael’s critiques of the AI industry extend beyond Anthropic, especially as other companies including OpenAI call for the government to create new regulations. He said that his background as an Uber executive—where he was chief business officer from 2013 to 2017—taught him how to take a more balanced approach to the fears that come with rapid technological progress. Still, Michael argued the frontier labs in their calls for stricter regulation on safety are acting like drunk drivers who demand someone pull them over instead of doing it themselves. “That kind of thing doesn’t make logical sense to me,” Michael said.

He did acknowledge escalating national security risks with AI. “With Chinese open source becoming dominant in the U.S., I worry about that a little bit because those companies are subject to the rule of the Chinese government,” Michael said. One option for checking the growing use of those models in the U.S. would be auditing the underlying weights of the software. But Michael said he preferred that the U.S. address the challenge with the growth of U.S. open-source options such as Reflection, Thinking Machines, and Nvidia’s Nemotron, with the understanding that none of those companies are at the frontier yet.

Michael said another worry is corporate espionage, such as the recent case of a Saudi employee of OpenAI installing a spying device. “Obviously all foreign nationals aren’t bad, but foreign nationals that have other loyalties bring with them some risks,” Michael said. He argued that rather than further government controls, companies should be more vigilant. He pointed to a service where the FBI will vet any foreign national employee if a company is worried about their background.

Aside from Michael’s increasing AI responsibilities within the Trump administration, he also helps spearhead Defense Department investment and modernization opportunities. That has included financing through the Office of Strategic Capital, originally created under the Biden administration to support lending to companies that strengthen defense supply chains.

The Trump administration has already announced hundreds of millions of dollars worth of loans to drone, rare earth and critical mineral companies. Michael said that the Defense Department is also looking at microelectronics companies doing different parts of the supply chain currently done by ASML and Taiwan Semiconductor Manufacturing Co., as well as drone parts where the U.S. is dependent on adversary nations, such as brushless motors and batteries. “You’ll see a steady drumbeat of continued deals from us,” he said.

Despite Michael’s central role in the Defense Department’s tech strategy, he said he did not have a view into one of its most scrutinized controversies under the Trump administration: Whether AI-enabled software failures from Palantir contributed to a U.S. missile strike on Iran in February that killed more than 150 people, including at least 123 children. “I haven’t read what you might have read, and I certainly haven’t read any report,” Michael said. “I’m not as online as I used to be.”

Leo Schwartz is a reporter covering the intersection of tech and politics. He can be reached at [email protected] or on Signal at leomschwartz.24

Amir Efrati is executive editor at The Information, which he helped to launch in 2013. Previously he spent nine years as a reporter at the Wall Street Journal, reporting on white-collar crime and later about technology. He can be reached at [email protected] and is on X @amir