Quick Answer
The Justice Department has urged a federal judge to reject the claim that training large language models on copyrighted text automatically violates copyright law. The September 1 filing supports OpenAI’s fair-use position, but it is not a ruling and does not resolve disputes over how training data was obtained or whether model outputs infringe. Creators and AI companies face a court decision.
Key Takeaways
- The Justice Department filed a Statement of Interest on September 1, 2026, in the OpenAI copyright litigation.
- The filing urged the court to reject the argument that LLM training on copyrighted text is inherently unlawful.
- The government described AI training as extraordinarily transformative, while separating training from output and data-acquisition questions.
- The filing is advisory, and Judge Sidney H. Stein will decide the underlying copyright issues.
- The New York Times argues that AI companies should pay fairly for material used to build their products.
What did the Justice Department tell the court about AI training fair use?
The Justice Department urged the U.S. District Court for the Southern District of New York to reject the argument that training large language models on copyrighted text violates copyright law. The government submitted its Statement of Interest on September 1, 2026, in In re: OpenAI Inc. Copyright Infringement Litigation, docket number 25-md-03143. The court-filed Statement of Interest sets out the government’s position.
The Justice Department’s position supports OpenAI’s argument that AI training can qualify as fair use, a copyright doctrine that can permit unlicensed use in particular circumstances. The government did not tell the court that every use of copyrighted material in AI development is legal. Instead, the filing argues against treating the act of training an LLM as a categorical copyright violation.
This distinction matters because the Times case is part of a wider set of lawsuits brought by authors, publishers, music labels, and news organizations. The filing appears to be the first time the U.S. government has formally weighed in on the wave of AI-training copyright cases. Reuters’ account of the filing describes the government’s intervention and its broader policy argument.
Why does the Justice Department consider AI training transformative?
The Justice Department described AI training as “extraordinarily” transformative because the government views the process as creating systems that generate and analyze language in new ways rather than simply republishing the source material. The filing says restricting that process would hinder scientific progress, prosperity, and economic mobility. That argument goes to a central fair-use question: whether a new use has a meaning or purpose distinct from the original work.
The government also argued that mandatory licensing for AI training would create a barrier to entry. In the Justice Department’s view, a licensing requirement could disproportionately benefit legacy media companies because they hold large volumes of written content. The practical implication is that the government sees broad licensing obligations as a market-structure issue as well as a copyright issue.
AI training fair use remains a legal argument, not a settled rule for every model or dataset. Courts examine fair use through fact-specific questions, and the government’s filing does not remove the need to evaluate the particular conduct alleged in the OpenAI litigation. Readers following newer AI products, including complex computer-task models, should separate product capability claims from the unresolved legal questions surrounding the materials used to develop such systems.
What does the DOJ filing not decide about copyrighted AI training?
The Justice Department filing does not decide whether OpenAI or any other AI company wins its copyright case. Judge Sidney H. Stein remains responsible for deciding the underlying questions in the consolidated litigation. The government’s submission is advisory, which means it gives the court the federal government’s legal and policy position without replacing the judge’s analysis.
The Justice Department also separated LLM training from two related but distinct issues: how training data was acquired and whether a model’s outputs infringe copyrighted works. That separation is important because a court could assess the training process differently from allegations involving unauthorized access, copying practices, or outputs that closely reproduce protected content.
Questions about how AI systems obtain material can carry legal and security implications beyond the fair-use debate. For example, concerns over AI agent access issues involve a different set of facts from the copyright analysis of a model learning statistical patterns from text. The sensible interpretation is that the DOJ filing narrows its argument to training, rather than offering a blanket defense for all conduct connected to AI development.
| Question in the dispute | Justice Department position | What remains unresolved |
|---|---|---|
| Training an LLM on copyrighted text | The government urged the court not to treat training as automatically unlawful. | Whether fair use applies to the conduct alleged in the case. |
| Obtaining training data | The government treated acquisition as a separate issue. | Whether particular acquisition methods create liability. |
| Model outputs | The government treated outputs as separate from training. | Whether particular outputs infringe protected works. |
| Compensation for creators | The filing opposed mandatory licensing as a general solution. | Whether courts or lawmakers impose obligations in future cases. |
How does the DOJ position affect the New York Times lawsuit?
The Justice Department’s position gives OpenAI support from the federal government in a case brought by The New York Times in 2023. The Times alleges that OpenAI and Microsoft used millions of newspaper articles without permission to train ChatGPT and related models. The filing does not end that lawsuit, but it gives the court a detailed argument for rejecting the plaintiffs’ broad theory about AI training.
The DOJ intervention matters because federal agencies can influence how courts understand the public-interest consequences of a legal rule. The government argued that the public and creative benefits of LLM training outweigh competitive harm, while distinguishing those claims from disputes involving outputs and data acquisition. The Associated Press reported that the Justice Department framed those issues separately in its submission.
The New York Times lawsuit remains focused on allegations that must be tested in court. A favorable DOJ position does not establish that the judge will accept OpenAI’s fair-use argument, and an unfavorable ruling would not automatically decide every other AI copyright case. For readers, the immediate effect is legal significance rather than a change to how ChatGPT or related products work.
Why are creators challenging AI training fair use?
Creators are challenging AI training fair use because they argue that AI companies should not be allowed to use protected work to develop commercial systems without permission or payment. The New York Times said the administration was siding with major AI companies over creators, and spokesperson Graham James argued that AI companies should pay fairly for material used to build their products.
The disagreement is not only about whether an AI model stores or reproduces an article. Creators and publishers also raise concerns about market substitution, bargaining power, and the value of work used at scale. The Justice Department’s filing directly addresses that economic debate by warning that mandatory licensing could favor organizations with the largest text libraries.
Other litigation shows that AI-training questions extend beyond newspapers and books. A streamer training lawsuit similarly reflects disputes over whether platforms can use creator material to train AI systems. The cases do not necessarily share the same facts or legal claims, so readers should avoid assuming that a government filing in the Times litigation determines the result elsewhere.
What could the court’s decision mean for AI companies and publishers?
The court’s decision could shape how AI companies and publishers assess copyright risk in future training projects. A ruling that accepts OpenAI’s argument could strengthen the position that training on copyrighted text can be fair use when the process is sufficiently transformative. A ruling against OpenAI could increase pressure for licenses, negotiated agreements, or changes to how companies assemble training data.
The immediate limitation is that a district court decision would address the claims and record before Judge Stein. Other courts could reach different conclusions based on different facts, including the type of material involved, the method of obtaining it, and evidence about market effects. The broader legal framework will likely develop through additional cases and possible appeals.
Publishers and creators should not treat the DOJ filing as a final answer about compensation, consent, or enforcement. AI companies should not treat the filing as permission to ignore disputes about data acquisition or output behavior. Organizations with material or products at stake should consult qualified copyright counsel before making licensing, data-use, or litigation decisions.
What should AI users make of the DOJ’s fair-use argument?
AI users should understand that the Justice Department’s fair-use argument concerns the legal status of model training, not a guarantee that AI outputs are accurate, safe, or free to reuse. The government explicitly treated outputs as a separate issue. That means a user who receives text, images, code, or summaries from an AI system still needs to consider accuracy, confidentiality, and potential copyright concerns.
Businesses should keep sensitive information out of public AI tools unless their contracts and internal policies permit that use. Copyright litigation does not change the practical need to review AI-generated work before publishing it, submitting it, or relying on it for a commercial decision. The same caution applies to systems that automate tasks or access third-party services.
For most consumers, the filing is best understood as an important legal development rather than a change in daily AI use. The practical response is to follow the case, read product terms carefully, and avoid treating an AI-generated output as automatically lawful to copy or distribute. The court’s eventual ruling will matter more than the government’s advisory filing for the parties in this case.
FAQ
Did the DOJ rule that AI training is fair use?
No, the Justice Department did not rule that AI training is fair use. The September 1, 2026, filing urged the court to reject a broad argument against LLM training, but Judge Sidney H. Stein will decide the copyright questions in the OpenAI litigation.
Does the DOJ filing mean OpenAI has won the New York Times case?
No, the DOJ filing does not mean OpenAI has won the New York Times case. The government’s Statement of Interest is advisory, and the court must still evaluate the parties’ claims and the relevant copyright law.
What does transformative use mean in the DOJ’s AI argument?
Transformative use means the government views LLM training as serving a new purpose rather than simply republishing the copyrighted text. The Justice Department described AI training as extraordinarily transformative, but the court must decide how that argument applies to the facts of the case.
Does the DOJ position cover AI outputs that copy copyrighted work?
No, the DOJ position treats AI outputs as a separate issue from model training. A model output that allegedly reproduces protected material can raise different legal questions from the use of text during training.
Should publishers and creators expect payment for AI training data?
No, the DOJ filing does not establish a right to payment for AI training data. The government argued that mandatory licensing could create barriers to entry, while the New York Times argues that AI companies should pay fairly for material used to build their products.