AI, Fair Use, and the Arsenal of Democracy

National security has been barely a whisper in the rollicking debates on AI rules and regulations. However, more and more defense programs are tapping directly into the products of what might be termed the “AI industrial base.”  From Air Force battle management to Navy shipbuilding and Army sensors and offensive cyber ops, AI is becoming an important element in U.S. multi-domain dominance.

As every indicator shows, the struggle with China is increasingly focused on AI. U.S. military programs are counting on the AI lead over China established by American technology companies.

That’s why two California court rulings on AI models and the Fair Use doctrine – seemingly just a boisterous Hollywood issue – have implications for national security, too.

Case two was brought by 13 authors against Meta, the purveyor of Llama AI. Here, Judge Vince Chhabria found that training AI models on purchased books did not infringe copyright because the use was “highly transformative.”  In other words, the trained AI model's end product bore no resemblance to the copyrighted novel or play.

Hollywood squawked. “Bad News for Movie Studios” headlined the Hollywood Reporter, posting an image of Darth Vader from Star Wars. 

The White House cheered. “Positive ruling for AI. There must be a fair use concept for training data or models would be crippled. China is going to train on all the data regardless, so without fair use, the U.S. would lose the AI race,” White House AI Czar David Sacks posted on X.

It’s not Hollywood’s job to factor in national security. Discussions around AI and creativity will persist. However, a new key issue is indeed emerging: allowing American AI models to continue training on the highest-quality data is crucial to maintaining the lead over China.

Granted, today’s AI industrial base bears little resemblance to its 20th-century counterparts among the shipyards and aerospace factories. During the Cold War, the military was the top customer for innovation. The scale of defense spending meant that it was government money, often from the Department of Defense, which drove the research agenda and paid for the development of advanced aircraft, ships, missiles, and satellites.

With AI, the investment position is completely different. The arrival of ChatGPT in 2022 supercharged the race against China for the global lead in AI. Government money is not leading the way. Instead, tech companies invest far more in research and development for AI than the government, and the amount of money is staggering. In the United States, private investment in AI increased from approximately $20 billion in 2018 to over $109 billion in 2024, as tracked by the Stanford HAI.  For 2025, Meta, Microsoft, Amazon, and Alphabet planned $325 billion in capital investments to build out AI, representing a 46% increase from 2024.  Obviously, there is no way for the U.S. government to replace or take over the level of funding needed to advance AI.

The AI arsenal of democracy thrives on intense competition among AI leaders. A strong business climate within the U.S. is the engine driving growth.

This is why the “Fair Use” precedents established by the California decisions matter for national security. Teaching large language models to generate human-like answers, dialogue, and essays requires extensive practice with written work. Large language models like ChatGPT2, Claude2, and Llama2 were trained using “a large corpus of high-quality data.”  The qualitative edge is crucial. Just imagine: AI models banned from the latest U.S. fiction but feasting away on Marx and Lenin and the latest “Xi Jinping thought” from China.

Even with hefty investment, America’s AI lead is in jeopardy. China’s startling DeepSeek chatbot has narrowed the gap in AI model capabilities, and a recent RAND study projected that China would close the gap in AI model benchmarks this year. 

In April, Xi Jinping led a Politburo session to revamp China’s AI policy – the first of its kind since 2018. “We must use AI to lead a paradigm shift in scientific research and speed up S&T innovation and breakthroughs in every field,” Xi exhorted. Johanna Costigian of the Paulsen Institute noted that Xi commanded Chinese researchers to make the next “paradigm shift” in AI capability.  “It’s game on for AI supremacy,” as China scholar Paul Triolo summed up, commenting that “Chinese leadership likely sees an existential threat from falling behind on AI development.” 

Fortunately, the U.S. still holds major strategic advantages. Export controls on highly advanced chips have restricted China’s options, despite the TSMC breach, which involved sales of advanced chips to a Chinese proxy. U.S. compute capacity is estimated to be ten times that of China.  Beyond this is the U.S. ability to experiment and innovate. Training in the full canon of literature, especially fiction, is part of retaining the qualitative advantage for the AI industrial base.

Already, a wide array of military programs are tapping into this same AI “industrial base.”  The Army has an AI accelerator called Operation Linchpin, where innovative companies of different sizes compete to improve speed and accuracy with AI for the Army’s battlefield sensors.  The Army’s goal is to establish an AI ecosystem that begins with secure, trusted data, AI model training, verification, and rapid deployment to Soldiers at command posts and in vehicles.

At Nellis Air Force Base in Nevada, an Air Force experiment with incorporating AI tools in wargame battle management “doubled the number of operational ‘dilemmas’ humans were able to address and generated three times as many ‘valid solutions’ to those problems, ranging from striking an enemy target to refueling friendly planes,” according to the outbrief.   The net effect was to “dramatically reduce decision time and improve decision quality for air battle managers working in complex operational environments,” said Colonel Christopher Cannon, who was the Advanced Battle Management Cross-Functional Team lead.

AI can unify threat intelligence data sets, tracking Russia’s shadow fleet and undersea cable saboteurs in the Baltic Sea, for example.  Anduril teamed with Meta on May 29 “to design, build, and field a range of integrated XR products that provide warfighters with enhanced perception and enable intuitive control of autonomous platforms on the battlefield.”

There’s even a case where AI meets the classic industrial base. Nuclear shipbuilder HII announced a partnership with C3 AI at the shipyard “to leverage data and digital capabilities like artificial intelligence in the urgent work of delivering ships to the U.S. Navy,” said HII CEO Chris Kastner. 

It should go without saying that AI has already been an integral part of stealth aircraft operations, as seen in the F-22, F-35, and B-2's strikes on Iran.

Boisterous debates on AI will continue. However, to ensure the U.S. military can tap into leading-edge AI, the first step is to keep America’s AI industrial base ahead of China's. National security deserves a voice in AI laws and regulations.

Speaking of Darth Vader, Amazon MGM Studios recently announced Spaceballs 2.  And no, it’s not a copyright infringement. It’s a parody and the direct rip-off of Star Wars characters is protected. “After 40 years, we asked what do the fans want,” said director and producer Mel Brooks. “Instead, we’re making this movie.”  AI still has a long way to go before it conquers Hollywood.

Dr. Rebecca Grant is a national security analyst based in Washington, DC specializing in defense and aerospace research and national security consulting.