Cold War-Era Defense Production Act Could Give Trump Leverage Over AI Companies Before December Expiration
The Defense Production Act, a Korean War-era law, could provide the executive branch with significant authority over frontier AI companies if their systems pose national security risks. With the law set to expire in December, experts debate how existing authorities could be used to demand information from AI developers.
The Defense Production Act, a Cold War-era law enacted in 1950 to mobilize American industry during the Korean War, could become a key tool for President Donald Trump to assert control over artificial intelligence companies if their systems ever threaten national security, according to national security and legal experts. The law's authorities, which include powers over federal contracts, exports, and civil liability, are set to expire in December, setting up a reauthorization fight later this year.
Congress recently extended most of the DPA's authorities, which were originally scheduled to lapse on September 30, through December 11, 2026. The extension came after leaders at some of the nation's most powerful AI companies issued stark warnings about increasingly capable models and backed calls to slow development so safety measures can catch up. Trump said this week that his administration already has «tremendous criminal and regulatory power» over AI companies, pushing back on industry calls for new guardrails.
James Lewis, director of the Strategic Technologies Program at the Center for Strategic and International Studies, said the authorities could be clearer but that a solution could likely be assembled from existing legal tools. He noted that the cyber threat landscape has changed dramatically, with vulnerabilities that once took months to find now taking days or hours, and AI helping attackers chain smaller weaknesses into more damaging attacks.
«The threat itself is not new,» Lewis said, recalling a recent conversation with a senior official from a Five Eyes intelligence partner. «We've known about this for a while, but people weren't taking it seriously.» AI, he said, has «changed the landscape dramatically.»
The Trump administration has already begun building a framework to identify when AI capabilities become dangerous. A June executive order directed federal agencies, including the NSA and CISA, to establish classified benchmarks for advanced cyber capabilities and determine when an AI system should be designated a «covered frontier model.» The order also calls for a voluntary framework through which developers can give the government early access to covered models before release, while stopping short of mandatory federal licensing or pre-clearance for frontier AI.
But knowing a dangerous capability exists is only the first problem. The next question is what Washington can actually do about it. One potentially powerful use of the DPA would come before an AI system ever caused damage: forcing companies to tell the government what they are developing. James E. Baker, a former chief judge of the U.S. Court of Appeals for the Armed Forces who previously served as a legal adviser to the National Security Council, pointed to the DPA's broad information-gathering authority.
«What you could use the DPA for is to require reporting, and that's the first step to making sound policy,» Baker said. «You can have a debate about whether there should be regulation and what the regulation should look like. But you really can't argue against knowing what's happening.» He added that knowing what is happening is the first place to start, after which the government can react with a policy prescription, a legal prescription, or nothing at all.
The DPA's limits, however, are just as important. Paul Rosenzweig, a lawyer and former deputy assistant secretary for policy at the Department of Homeland Security, said the law was principally designed to ensure the government could obtain goods and services needed for national defense, not to give Washington broad power to stop private companies from producing them.
«The Defense Production Act is more about mandating the production of things for the United States rather than preventing the production for the general public,» Rosenzweig said. Where it could prove particularly useful, he agreed, is giving the government access to information. «What it does give them, I think, and this is the best part about it, is it does give them a pretty good handle for demanding more information about a defense-sensitive kind of product, which frontier AI certainly could be reasonably characterized as,» he said.
If the DPA could help Washington understand what AI companies are building, other existing authorities could potentially come into play depending on the threat. Export controls can restrict sensitive technology from reaching foreign adversaries. Government contracting rules can impose conditions on companies seeking federal business. And criminal laws continue to apply when AI is used to commit crimes. The June executive order, for example, directed the attorney general to prioritize enforcement of existing federal criminal laws against people who use AI to commit offenses.
Exactly how the government could use the DPA's powers against a dangerous AI system remains largely untested. But with the law's expiration looming in December, the debate over how to oversee frontier AI is likely to intensify in Washington.
