The fight over whether AI companies can train on copyrighted work is moving beyond a dispute between creators and technology companies. The U.S. government is now arguing that access to enormous amounts of human-created knowledge may also affect national competitiveness, scientific progress and national security.
WHAT’S HAPPENING
The U.S. Justice Department has intervened in The New York Times’ copyright lawsuit against OpenAI and Microsoft, supporting the argument that training artificial-intelligence models on copyrighted written material can generally qualify as fair use.
The government told the federal court that AI training is highly transformative and warned that interpreting copyright law too restrictively could hinder American AI development, scientific progress and economic competitiveness.
The filing does not settle the legal issue. Courts will ultimately determine how copyright law applies to different forms of AI training.
At the same time, Commerce Secretary Howard Lutnick urged G20 governments to embrace frameworks that allow AI companies to train models on creators’ work while also finding ways to protect artists and creators.
WHY IT MATTERS
This changes the scale of the copyright debate.
Until now, much of the argument has sounded like:
Did an AI company have the right to use this book, article, image or song?
The emerging government argument introduces another question:
What happens to a country’s AI capabilities if its models cannot learn from large portions of recorded human knowledge?
That turns copyright into something larger than a fight over licensing.
Books, journalism, research, art and other human-created material are becoming part of the raw informational environment from which increasingly capable AI systems are built.
The United States is effectively signaling that access to that environment may carry strategic value.
That does not eliminate creators’ property rights.
But it creates a collision between two legitimate interests:
Protecting the people who produce valuable knowledge — and ensuring AI systems can learn from enough knowledge to remain useful and globally competitive.
WHO BENEFITS
AI developers. A broad interpretation of fair use could dramatically reduce the need to negotiate licenses for every piece of material used during model training.
Smaller AI companies. Licensing enormous training libraries could favor the wealthiest technology companies. Broad fair-use protections could lower one major barrier to entering the AI market.
The U.S. AI sector. The government’s argument explicitly connects AI development with economic competitiveness, scientific progress and national security.
WHO LOSES
Publishers, writers, artists and other copyright owners could lose negotiating power if courts conclude that substantial amounts of copyrighted material can legally be used for AI training without permission or payment.
There is also a broader economic problem.
If AI systems derive value from human-created work while the people and institutions producing that work receive progressively less compensation, the system could eventually weaken the very ecosystem generating the knowledge AI depends upon.
WHAT HAPPENS NEXT
The courts still have to draw the line.
Different judges have already reached differing conclusions in earlier AI copyright disputes, and the legality of training can depend on factors including how material was obtained, how it was used and whether the resulting system competes with the original works.
But something larger has now changed.
The federal government is no longer treating AI copyright solely as a private dispute between technology companies and creators.
It is beginning to frame access to training material as part of America’s ability to build competitive artificial intelligence.
That could create one of the defining policy questions of the AI era:
When human knowledge becomes an input into national technological capability, where does intellectual property end — and infrastructure begin?
