Defense Secretary Pete Hegseth has delivered a stark ultimatum to artificial intelligence company Anthropic this week, demanding that it opens its AI technology for unrestricted military use by Friday or faces the loss of its valuable government contract. This move highlights escalating tensions between the Pentagon and tech firms over the ethical deployment of artificial intelligence in national security operations.
Unprecedented Use of Defense Production Act
Defense officials within the Trump administration have warned that they could designate Anthropic, the creator of the AI chatbot Claude, as a supply chain risk. Alternatively, they might invoke the Cold War-era Defense Production Act to grant the military sweeping authority to utilise Anthropic's products, even without the company's approval. Such an application of the law would be unprecedented and could provoke significant legal challenges, according to experts.
The government's aggressive stance underscores a broader, contentious debate regarding artificial intelligence's role in national defence. Anthropic stands as the last major AI firm refusing to supply its technology to a new U.S. military internal network. CEO Dario Amodei has consistently voiced ethical concerns about unchecked government use of AI, including risks associated with fully autonomous armed drones and AI-assisted mass surveillance that could monitor dissent.
What is the Defense Production Act?
The Defense Production Act, signed by President Harry S. Truman in 1950 during the Korean War, provides the federal government with broad authority to direct private companies to meet national defence needs. Over decades, its powers have been invoked not only in wartime but also for domestic emergency preparedness, recovery from terrorist attacks, and responses to natural disasters.
One key provision allows the president to require companies to prioritise government contracts deemed essential for national defence, ensuring the private sector produces sufficient goods for war efforts or emergencies. Other clauses enable the president to use loans and incentives to boost production of critical goods and authorise voluntary agreements with private industry.
Joel Dodge, an attorney and director of industrial policy at the Vanderbilt Policy Accelerator, describes the DPA as "one of the government’s most powerful and adaptable industrial policy tools." However, he notes that using it to compel a company to produce a product it deems unsafe or to dictate terms of service would be without precedent in the law's history.
Historical Applications of the DPA
Former President Donald Trump and President Joe Biden previously invoked the DPA to enhance supplies during the COVID-19 pandemic. Biden also used it during the 2022 nationwide baby formula shortage to accelerate production and authorise import flights. In 2023, Biden employed the act in an executive order on AI, requiring companies to share safety test results with the government, though Trump repealed this order at the start of his second term.
Historically, the administrations of Presidents Bill Clinton and George W. Bush utilised the DPA to ensure electricity and natural gas supplies to California utilities during an energy crisis. After Hurricane Maria struck Puerto Rico in 2017, the law prioritised contracts for food, water, housing, and electrical system restoration.
The DPA requires periodic reauthorisation to remain effective, with its next expiration set for September 30 this year. Depending on how the Pentagon's demands progress, Anthropic could become a focal point for lawmakers during reauthorisation discussions.
Potential Consequences and Legal Battles
If the Defense Department invokes the DPA provision to prioritise government contracts and order production of specific goods, Anthropic could push back if the requested product is not something it already produces or if the terms are deemed unreasonable. However, the government may attempt to overrule such objections.
Charlie Bullock, a senior research fellow at the Institute for Law & AI, suggests that if neither side concedes, litigation between Anthropic and the government is a realistic outcome. There is also noted tension in the Pentagon's dual warnings: designating Anthropic as a supply chain risk while simultaneously asserting that its products are crucial enough to national defence to warrant DPA invocation.
Dodge believes the administration is counting on various pressures to compel Anthropic to acquiesce by Friday. In potential future litigation over a DPA order, he doubts the government would prevail, viewing such action as "very out of bounds under the text of the law." However, if successful, this could open a "Pandora’s box" regarding government power over private companies.
The outcome of this standoff will significantly influence the future relationship between the tech industry and national security agencies, setting a precedent for how ethical considerations in AI development are balanced against defence imperatives.
