Elias unpacks the research behind the headlines in plain language.
Get the week's top AI and AI-art stories delivered to your inbox — curated, concise, free.
Free. Unsubscribe any time.
Pick a companion and get their take on this story
The Trump administration has filed a legal brief in The New York Times' copyright lawsuit against OpenAI, arguing that training large language models on copyrighted material is in the national interest — a position that, if it shapes court precedent, would directly affect every AI image generator trained on scraped visual data.
The Department of Justice brief, cited by TechCrunch, frames AI training as a matter of geopolitical competition rather than purely a copyright question.
"The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally."
— DOJ Amicus Brief
The administration filed as an amicus — literally "friend of the court," meaning a party that is not directly involved in the lawsuit but submits arguments to inform the judge's reasoning. Amicus briefs do not decide cases, but a government brief carries weight because courts treat the executive branch's policy views as relevant context when interpreting law.
The NYT lawsuit, filed in December 2023, alleges OpenAI unlawfully trained its systems on Times articles without permission or payment. The Times is seeking what The Verge describes as "billions of dollars" in damages.
The case is nominally about text, but the underlying legal question — whether scraping and training on copyrighted work qualifies as fair use — applies equally to image diffusion models trained on billions of unlicensed photographs and illustrations. A ruling that fair use does not cover AI training would expose Stable Diffusion, Midjourney, and similar systems to the same legal theory the Times is using against OpenAI.
Fair use is a legal doctrine that permits limited use of copyrighted material without permission under certain conditions, including whether the new use is "transformative." AI companies have consistently argued that training is transformative; rights holders consistently argue it is not. No federal appellate court has settled the question.
The government's intervention tilts the policy environment toward the AI industry's preferred reading. Whether a district court judge finds that persuasive is a separate matter — courts are not bound by the executive branch's preferred outcome.
The DOJ brief arrives while Sony Music and Warner Chappell are pursuing a separate copyright suit against Anthropic, seeking up to $150,000 per copyrighted work — a case Charmloop covered in detail in our earlier report on Sony Music and Warner's suit against Anthropic. Internal Anthropic communications surfaced in that case have already complicated the industry's public messaging around training data practices.
Google, meanwhile, has been negotiating licensing deals with Hollywood studios rather than litigating, a strategy that implicitly acknowledges the legal risk of the scrape-first approach — as Charmloop reported in Google's Hollywood licensing push.
For anyone using AI image generation tools on platforms like Charmloop's image generator or browsing the model catalog, the practical stakes are straightforward: if courts ultimately rule that training on unlicensed data is infringement, the models themselves could face injunctions, forced retraining, or licensing costs that reshape which tools remain available and at what price.
The government's brief does not resolve that risk — it signals which direction the current administration wants the law to go. The actual precedent will come from judges, and the NYT case is still in early stages. Creators who want to stay ahead of how these legal developments affect model availability and training practices will find the next few court filings worth tracking closely.
