Trump Administration's Cold War-Era Tool Could Address AI Security Threats

A powerful Cold War-era tool, the Defense Production Act (DPA), could be utilized by the U.S. executive branch to address potential national security threats posed by artificial intelligence, according to national security and legal experts. Originally enacted in 1950 to mobilize American industry during the Korean War, the DPA grants authorities to expand military production, prioritize government contracts, and respond to national emergencies.

While the exact application of DPA powers against a dangerous AI system remains largely untested, experts suggest existing authorities within the act that cover cyberattacks, exports, federal contracts, and civil liability could be adapted. James Lewis, director of the Strategic Technologies Program at the Center for Strategic and International Studies, noted that while the authorities could be clearer, a solution could likely be assembled using current legal frameworks.

The reauthorization of most of the DPA's authorities in September 2026, through December 11, 2026, has intensified discussions about its potential role in governing AI. This extension sets the stage for further debate on the act's future, particularly in the period following the November midterms.

Recent warnings from leaders at major AI companies about the escalating capabilities of AI models and calls to slow development have added urgency to these discussions. Former President Donald Trump has stated that the administration already possesses significant regulatory and criminal power over AI companies, pushing back against calls for new guardrails.

Lewis highlighted the evolving nature of cyber threats, where AI can accelerate the discovery of vulnerabilities and enable attackers to combine smaller weaknesses into more damaging attacks. He noted that while the threat is not new, AI has dramatically altered the landscape.

The Trump administration has begun establishing a framework to identify dangerous AI capabilities. A June executive order directed federal agencies to create benchmarks for advanced cyber capabilities and determine when an AI system should be designated a "covered frontier model." The order also proposed a voluntary framework for developers to provide early access to these models, though it stopped short of mandatory federal licensing.

James E. Baker, former chief judge of the U.S. Court of Appeals for the Armed Forces, pointed to the DPA's broad information-gathering authority as a key component. "What you could use the DPA for is to require reporting, and that's the first step to making sound policy," Baker told Fox News Digital. "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."

However, legal experts also note the DPA's limitations. Paul Rosenzweig, a lawyer and former deputy assistant secretary for policy at the Department of Homeland Security, stated that the act was primarily designed to ensure the government could obtain necessary goods and services for national defense, rather than to prevent private companies from producing certain items. "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.

Rosenzweig agreed that the DPA could be particularly useful in demanding more information about defense-sensitive products, which frontier AI could arguably be characterized as. Beyond the DPA, other existing authorities could be applied depending on the specific threat. Export controls can restrict the transfer of sensitive technology, government contracting rules can impose conditions on companies, and criminal laws remain applicable when AI is used to commit crimes.

The June executive order, for instance, directed the attorney general to prioritize enforcement of existing federal criminal laws against individuals using AI to illegally access or damage computer systems. Trump has also directed national security agencies to expand their use of frontier AI while ensuring military and intelligence systems remain controllable and maintain clear lines of human accountability.

Rosenzweig emphasized that there is no broad AI exemption from generally applicable laws, and a range of existing legal authorities, from tort liability and privacy rules to consumer protections and criminal law, could potentially apply to frontier AI. However, he cautioned that the effectiveness of this legal patchwork against a true AI emergency remains untested.

He added that if an AI system were to cause damage autonomously, determining developer negligence or applying stricter product liability standards would present challenges in court. "I tend to think that liability is going to be difficult to establish early on," Rosenzweig said. "Right now, that's not where I would base my deterrence model."

This uncertainty highlights the limitations of the government's current toolkit, as many laws have not been tested against systems exhibiting increasing autonomy. For now, policymakers are applying laws designed for different threats to a rapidly evolving technology.