1950s Law Gives White House Leverage over AI Risks

Washington possesses a legal instrument from the Cold War era that could be deployed against artificial intelligence if the technology is deemed a threat to national security. The Defense Production Act, originally passed in 1950 to mobilize industry during the Korean War, grants the executive branch broad powers to direct resources and prioritize federal contracts. While previously used to manage military supply chains, these authorities now extend to cyber threats and critical technologies, providing a pathway for the government to intervene in the AI sector without new legislation.

The relevance of this law has grown as leaders at major AI companies warn that model capabilities are advancing faster than safety measures can keep up. President Trump has indicated the administration already holds significant regulatory power over these firms, pushing back on calls for new external guardrails. However, the legal basis for this stance rests on a temporary extension of the Act’s authorities, which Congress recently prolonged through December 2026. This creates a fragile window of leverage that will soon require another legislative fight, potentially coinciding with shifts in congressional control after the upcoming midterms.

Cold War Law Targets Modern Tech

Experts note that the specific application of these powers to AI remains largely untested in courts. The Act covers a wide range of activities, including exports, federal contracts, and civil liability, which legal scholars argue can be adapted to address catastrophic risks from frontier models. According to reports from GN technics/ai (en-US), the administration has begun building a framework to identify when AI capabilities become dangerous. A June executive order directed agencies like the NSA and CISA to establish classified benchmarks for advanced cyber capabilities, determining when a system should be designated a covered frontier model.

The proposed approach relies on a voluntary framework where developers provide early government access to models before release. This method deliberately avoids creating mandatory federal licensing or pre-clearance regimes. Instead, it leverages existing authorities to demand transparency and reporting from companies. This allows the government to gather information and assess risks without establishing a new bureaucratic layer of approval, a trade-off that prioritizes speed and existing legal channels over comprehensive regulation.

Urgency Drives Safety Debates

The pace of technological change has altered the threat landscape significantly. James Lewis, director of the Strategic Technologies Program at the Center for Strategic and International Studies, explained that vulnerabilities which once took months to find can now be discovered in days or hours. AI systems can chain smaller weaknesses into more damaging attacks, a capability that has changed how intelligence partners view the risk. Lewis noted that while the threat is not new, the speed at which AI amplifies it has forced a reassessment of previous assumptions about safety and control.

Legal experts suggest that the most immediate use of the Act would be to force companies to report on their developments. James E. Baker, a former chief judge of the U.S. Court of Appeals for the Armed Forces, pointed to the Act’s broad information-gathering authority as a tool for sound policy-making. By requiring reporting, the government can establish a baseline of understanding before deciding on further regulatory actions. This approach allows for a debate about regulation based on data rather than speculation, addressing the core concern that current oversight mechanisms are lagging behind the speed of innovation.

December Deadline Creates Political Tension

The current legal authority is not permanent. Congress extended the relevant provisions of the Defense Production Act through December 11, 2026, setting the stage for another reauthorization battle. This timing is politically sensitive, as it occurs just before the legislative landscape could shift following the November midterms. If the extension lapses or is modified, the executive branch may lose its most potent legal tool for intervening in the AI sector. This uncertainty adds pressure on lawmakers to decide whether to maintain, expand, or repeal these powers, with the outcome potentially reshaping the balance of power between Washington and the private tech industry.

The catch in this strategy is that it relies on the continued willingness of Congress to uphold a Cold War-era statute for a 21st-century technology. The Act was designed for mobilizing physical production, not for managing the abstract risks of artificial intelligence. While legal experts argue the authorities are broad enough to be adapted, the lack of clear precedents means any action could face legal challenges. The government is effectively using a blunt instrument for a precise problem, a trade-off that offers immediate leverage but carries the risk of overreach or ineffective enforcement if the legal boundaries are tested in court.