With every special interest in Washington wrapping policy priorities in the American flag, Big Tech companies are framing unauthorized use of copyrighted works as essential for beating China in artificial intelligence. America's intellectual property rights, deeply rooted in the Constitution, have become the latest proposed sacrifice at the altar of technological progress.
On July 28, Microsoft held a dinner with Beltway China hawks arguing that policymakers should allow AI training on copyrighted works without licenses or compensation to rights holders. Without this access, Microsoft warned, the U.S. risks losing the AI race and compromising national security.
Yet Microsoft's own track record raises questions about whether these companies truly prioritize U.S. interests. Recent reporting revealed that Microsoft used China-based engineers for years to maintain Defense Department and other federal systems with minimal U.S. supervision. Those systems were recently hacked by China, and Microsoft had failed to disclose key details about its reliance on Chinese engineers—what the DoD called a "breach of trust."
This pattern extends beyond Microsoft. Despite positioning themselves as patriotic champions, major tech companies maintain risky business relationships with China while simultaneously arguing they need unrestricted access to American creative works to compete with Beijing.
Setting aside questions of dual loyalty, the fundamental premise—that training AI on copyrighted works is essential for national security—lacks compelling evidence. Proponents argue that general-purpose large language models require copyrighted content as inputs for defense applications, and that licensing would be too slow or expensive, allowing China to gain advantage.
But this conflates different types of AI systems. Do autonomous weapons or intelligence analysis tools really require training on Mickey Mouse or James Patterson novels? A growing number of AI experts question whether simply ingesting more data leads to more useful AI. Many national security professionals have concluded that commercial LLMs often prove unsuitable for defense applications due to their tendency to hallucinate, their vulnerability to adversarial attacks, and their inability to understand specialized terminology.
Defense-oriented AI applications typically require fine-tuning on specialized datasets—decades of government satellite imagery and weather data rather than episodes of Yellowstone. If tech companies prioritized genuine security applications over short-term commercial interests, they would focus on making government datasets machine-readable for specific agency uses rather than demanding blanket access to private intellectual property.
Meanwhile, America's creative industries provide demonstrable national security value through soft power projection. As the Center for Strategic and International Studies notes, "Economic security translates to national security, and the cultural and creative economy exists at the intersection of these two areas." Taiwan leverages its creative industries to counter Chinese influence, just as American entertainment has fueled military recruitment and built cultural capital globally. The AI-generated "slop" proliferating on social media lacks this authentic cultural resonance. TikTok clickbait is no substitute for Top Gun.
Unauthorized training of copyrighted works threatens the economic viability of industries responsible for over 2.3 million U.S. jobs, over $229 billion in wages, and a $37 billion trade surplus. The Trump Administration has made clear that the country's AI growth path must be "pro-worker" and a "potent tool for job creation in the United States." A policy of blanket allowance of training on copyrighted works under a novel interpretation of the "fair use" doctrine would directly contradict those goals.
The solution lies not in choosing between creativity and AI innovation, but in developing robust licensing markets. Major tech companies and startups have already demonstrated they can obtain licenses for valuable content through various agreements. Recent deals between news organizations, book publishers, and AI companies show these markets can function effectively when companies choose to participate. If Spotify can license all the music it needs, why can't OpenAI?
Policymakers should resist calls for blanket exemptions or premature intervention in ongoing litigation. Instead, they should encourage the development of licensing frameworks that serve both creators and AI developers. This approach respects constitutional property rights while enabling legitimate AI advancement.
Tech companies naturally prefer to get all the benefits of human creativity without paying, but economic convenience cannot justify abandoning constitutional principles or undermining American soft power advantages that took decades to build.
Will Chamberlain is Senior Counsel for the Article III Project.
