The Department of Justice has filed a statement of interest in one of the most closely watched copyright lawsuits of the AI era, siding with OpenAI and Microsoft against news organizations including The New York Times.
The newspapers argue the tech companies violated copyright law by training AI systems on millions of articles without permission or payment. OpenAI and Microsoft counter that this training qualifies as “fair use,” a legal doctrine that permits limited use of copyrighted material without the owner’s consent.
The DOJ’s filing embraces a sweeping version of that fair-use argument, one that critics say would grant AI companies broad access to copyrighted material for commercial training without licensing fees or creator consent. That position marks a departure from what had been a more balanced administration stance on AI and intellectual property.
The department’s reasoning centers on cost: requiring licenses for training material, it argues, would burden AI developers, potentially squeezing out smaller companies and weakening U.S. competitiveness and national security in the AI race.
A Break From the Administration’s Own Framework
That argument sits uneasily alongside other administration statements. The White House’s official AI framework acknowledges there are valid arguments on both sides of the copyright question. The Carolina Principles for Emerging Technologies, adopted by G20 ministers at the Commerce Department’s recent Innovation Ministerial, likewise recognize that AI and copyright raise “complex legal and policy questions” and affirm the “essential role” of copyright protections for creators.
There’s also a competitive wrinkle the DOJ’s position doesn’t address: some AI startups have said that paying for training material actually gives them more reliable access, less legal exposure, and better-performing models. Companies have emerged specifically to aggregate copyrights and simplify licensing deals. Large, well-capitalized AI firms arguing against paying for material they use are, notably, among the companies best positioned to afford it — and doing so would help build the very licensing markets the industry says it needs.
Property-rights concerns cut against the DOJ’s reasoning as well. Even where national security is invoked, the government generally cannot disregard property rights outright; the Constitution requires just compensation when private property is taken for public use, and intellectual property is property. Allowing AI firms to use creators’ work without permission or payment would function much like letting a data center developer seize land without paying for it.
What’s at Stake for Creators
If courts adopt a fair-use standard this broad, it would weaken the incentive structure that encourages people to keep writing, filming, composing, and publishing in the first place. Licensing revenue from copyrighted works underpins more than $1 trillion in U.S. economic activity. A weaker copyright regime could mean fewer new books, articles, research papers, films and songs — which would eventually leave AI systems with a thinner, lower-quality pool of material to train on.
A Legislative Alternative
One bipartisan proposal offers a narrower path. The CLEAR Act, introduced in February by Sen. Adam Schiff, D-Calif., and Sen. John Curtis, R-Utah, would require AI developers to publicly disclose which copyrighted works were used to train each new generative-AI model.
The bill wouldn’t resolve whether any specific training practice counts as fair use. But it would give writers, artists and publishers the information needed to know whether their work was used — and to pursue compensation or other remedies if they believe their rights were violated.
Rather than pushing courts toward a blanket fair-use exemption for AI training, supporters of the CLEAR Act and similar measures argue policymakers should instead work toward a functioning market where AI firms can license copyrighted material efficiently while creators retain control over their own work.


Add comment