Twelve House Democrats sent a formal letter to Attorney General Todd Blanche on November 19 demanding that the Justice Department explain how it came to file a statement of interest on September 1, 2026, arguing that training AI on copyrighted material is generally fair use, in a copyright lawsuit brought by the New York Times against OpenAI and Microsoft, without first consulting the Copyright Office or the Patent and Trademark Office. The letter, led by Rep. Chellie Pingree of Maine, asks for all internal documents and any external communications from companies or organizations that influenced the filing. Democrats are set to take the House majority in January, converting this request into a potential subpoena if the department does not respond by the end of November.
The September Filing That No Career Attorneys Signed
Statements of interest are legal filings the Department of Justice uses to inform federal courts of the official position of the United States government on a question of law. They are not briefs for one party, but courts treat them as representing the considered view of the entire executive branch. The Axios investigation that first reported on the Democrats' letter found something unusual about the September 1 filing: unlike most statements of interest in complex litigation, it was not signed by career antitrust attorneys alongside senior DOJ officials. Career attorneys are the institutional continuity of the department, the people who have worked across administrations on IP policy and who understand the existing precedents. Their absence from the signature block was noticed by practitioners who reviewed the filing.
According to VitalLaw's coverage of the government's position, the DOJ argued that AI model training is generally fair use under copyright law, meaning that an AI company's ingestion of copyrighted content to train a statistical model constitutes a transformative use that the law does not prohibit. That view, if adopted by courts, would dramatically limit the liability exposure of every major AI laboratory in the country. The department filed the statement without issuing a public notice, without a press release, and without detailed explanation of its reasoning in the accompanying documentation.
The case in question, The New York Times Company v. OpenAI, Inc., was filed in the Southern District of New York in December 2023. The Times argues that OpenAI and Microsoft copied millions of its articles to train GPT-4 and subsequent models, that the resulting outputs reproduce its content closely enough to constitute infringement, and that this competition threatens the publisher's ability to fund its journalism. The legal stakes extend far beyond one publisher: more than two dozen separate copyright suits against AI companies are currently working through federal courts, and many will hinge on the same fair use question the DOJ decided to weigh in on without following its typical internal process.
Axios also reported that the September filing appeared inconsistent with an earlier position the administration had taken in a White House AI policy framework published on March 20, 2026, stating that courts, not the executive branch, should decide whether AI training constitutes fair use. The Democrats' letter cites this inconsistency as one reason they are demanding internal records. The department's switch from "let courts decide" to "here is what courts should decide" in six months, without public explanation, is at the core of their concern.
What the Twelve Democrats Are Asking For
The letter, led by Pingree and co-signed by Reps. Becca Balint of Vermont, Deborah Ross of North Carolina, Raja Krishnamoorthi of Illinois, Judy Chu of California, Jared Huffman of California, Ro Khanna of California, and five additional members, makes two specific requests of the department, according to reporting by Rallies.ai. First, it asks the DOJ to produce all draft statements and internal documents used to develop the September filing. Second, it asks the department to identify any external companies or organizations that provided legal analyses, briefing materials, or other communications about the statement before it was filed.
The second request is the more pointed one. The letter's underlying concern is that the department may have written its legal position in coordination with the AI companies that benefit from it, rather than developing that position through its own independent analysis. That concern would not only raise questions about bias. It could call into question whether the filing genuinely represents a legal position of the United States government or a favor to private interests dressed up in official language.
The lawmakers set a deadline at the end of November for the DOJ to respond. The timing is deliberate: Democrats are on track to take control of the House in January, which will give the new majority subpoena power and committee chairmanships. A voluntary response in the next two weeks carries the implicit message that stonewalling will simply convert a letter into a subpoena issued with the full force of the new majority. The choice the department makes before November 30 is therefore consequential regardless of what the response actually contains.
Axios also noted that prior reporting found the copyright-related agencies were surprised by the DOJ's September filing. If the agencies charged with administering the Copyright Act did not know the filing was coming, that calls into question the administrative process that generated it. Congressional oversight exists precisely to investigate whether agencies followed their own rules when making decisions with large legal and commercial consequences for private parties and content creators alike.
The Copyright Agencies That Were Left Out
The U.S. Copyright Office is a legislative branch agency housed under the Library of Congress, with statutory responsibility for administering the Copyright Act and advising Congress on copyright policy. It is not part of the executive branch. The U.S. Patent and Trademark Office is an executive branch agency within the Commerce Department, with specific expertise in IP law and policy. When the DOJ files a statement of interest in a copyright case, standard practice is to solicit input from both agencies so that the government's position reflects their expert views on the relevant statutes and precedents rather than a single department's reading of the law.
That consultation did not happen in this case. The Copyright Office's publicly stated position on AI training is more cautious than the DOJ's. The Office has said that AI training is fair use only "in some circumstances," a carefully limited formulation that acknowledges the standard four-factor fair use analysis requires a case-by-case evaluation. The four factors are: the purpose and character of the use, including whether it is commercial; the nature of the copyrighted work; the amount and substantiality of what was taken; and the effect on the market for the original work. Each of these factors points in different directions for different types of training data.
Courts have never ruled that mass ingestion of complete, commercially valuable articles for the purpose of building a commercial AI product is automatically transformative. The Copyright Office's nuanced framing reflects that reality. The DOJ's categorical claim that training is "generally" fair use sets aside this complexity in favor of a broad assertion that the courts, the expert agencies, and now several dozen members of Congress find legally unsupported on its face.
Legal scholars have noted that statements of interest in complex IP cases are supposed to help courts navigate difficult legal terrain, not shortcut it with categorical claims. A filing that does not engage with the four-factor test, and that was developed without the input of the agencies that specialize in that test, carries weaker persuasive authority than it appears on its face. That weakness is now documented in the congressional record, which is precisely where plaintiffs' lawyers will point when they respond to it in their own briefs.
Raskin, Nadler and the Brief That Called It a Bulldozer
The November 19 letter is not the first time Congress formally objected to the DOJ's position in this case. On October 30, three senior Democrats on the House Judiciary Committee, Reps. Jamie Raskin of Maryland, Jerrold Nadler of New York, and Henry C. "Hank" Johnson Jr. of Georgia, filed their own amicus brief in the New York Times case asking the judge to disregard the government's statement of interest, according to Bloomberg Law. They called the DOJ's proposed blanket fair use exception a "bulldozer" that disregards federal copyright law and would financially devastate creative industries and the American creative economy.
That filing was notable for a specific reason: members of Congress rarely file amicus briefs in active federal cases. When they do, it usually signals that they believe the executive branch has taken a legally untenable position that the legislature needs to correct before it hardens into precedent. The fact that the November 19 letter follows just three weeks later, with a substantially larger group of signatories and a formal document demand, suggests that the congressional pushback is building rather than dissipating.
The underlying case has also expanded in scope. The New York Times suit was filed in December 2023 as a single-publisher action, but it has since become the lead case in a broader fight over whether AI companies can build billion-dollar products by training on content they never paid for. Book authors, music publishers, photo agencies, and other news organizations have filed parallel suits in federal courts across multiple circuits. How the DOJ's statement influences the fair use analysis in these satellite cases will depend in part on whether the courts treat it as credible government guidance or as a compromised filing that cannot be taken at face value.
The case is now at the summary judgment stage on fair use. Both sides have submitted their legal arguments on the core question and are waiting for the Southern District to rule. The complexity of the issues and the volume of pending litigation make a ruling before January 2027 unlikely, though a ruling could come at any time. If the court schedules oral argument, that hearing date will serve as the clearest public signal of the judge's timeline.
Why the Procedural Fight Could Hurt OpenAI More Than It Helps
The bear case for OpenAI is not the most intuitive one. Critics argue that the DOJ's intervention, however welcome from the company's perspective in September, may have made its legal position weaker rather than stronger by November. A statement of interest that arrives under a cloud of procedural questions, that was not signed by career attorneys, and that is now the subject of a formal congressional inquiry carries less judicial weight than one that sailed through normal review without controversy. If OpenAI's lawyers are relying on the September filing as evidence of executive branch support for their fair use defense, that support is now publicly contested in a way that invites the court to scrutinize it.
There is also a legislative risk that the company's lobbyists are tracking carefully. Congressional Democrats have signaled they will launch investigations into DOJ conduct when they take the majority. An inquiry that starts as a document request can escalate into hearings, and hearings produce a legislative record. A bill that removes the fair use defense for AI training, or that requires licensing fees for content used in model training, is not a theoretical scenario: several European Union member states have pushed for exactly that framework under the DSM Directive's text and data mining provisions, and U.S. lawmakers are studying the European approach closely as they build their own legislative positions on AI and copyright.
However, the skeptics' view is that this congressional pushback will not reach its stated objective, because the new House majority will face a Senate still controlled by Republicans who are broadly sympathetic to the technology industry's position. Any legislation that places a material burden on AI development, whether through mandatory licensing or expanded copyright liability, will likely stall in the Senate even if it passes the House on a party-line vote. The DOJ's September statement, however procedurally questionable, may still be the last official word the executive branch says on this question before courts rule on the merits.
That Senate arithmetic may be right in the near term. However, the procedural questions raised by the November 19 letter create a vulnerability that persists regardless of which party controls the upper chamber. When OpenAI's lawyers invoke the September statement in summary judgment briefing, the plaintiffs will now be able to respond by noting that the filing was developed without input from the Copyright Office, that career attorneys were not among its signatories, and that Congress has formally questioned whether outside parties shaped its content. Each of those facts is a reason for a court to discount the weight of the government's position, independent of which way the political winds are blowing in Washington.
The November 30 Deadline and the January Reckoning
The near-term milestones are well-defined. The DOJ has until roughly November 30 to respond to the Democrats' document request. If it provides a substantive response, it may demonstrate that the September filing was a legitimate, if expedited, government action and that no outside parties shaped its content. If it stonewalls or offers only a procedural acknowledgment, the new House majority will begin drafting a subpoena shortly after convening in January, and the investigation will deepen into what is likely to be the most visible AI-related congressional inquiry of 2027.
The New York Times case is at summary judgment, meaning the judge has the core legal question before them. A ruling on fair use, whichever way it goes, will immediately affect the more than two dozen parallel copyright suits against AI companies. If the court rules that training on journalism archives is fair use, it validates the DOJ's position on its merits regardless of how that position was produced. If it rules against OpenAI and Microsoft, the DOJ's intervention will have failed on its own terms and will give the plaintiffs in related cases a powerful political argument about why the executive branch's statement should not be credited as neutral legal guidance.
On November 18, the Ninth Circuit heard oral argument in xAI v. Bonta, Elon Musk's challenge to California's AB 2013, which requires AI developers to disclose the datasets used to train their models. That case tests a different but related question: whether states can mandate transparency about training data in the absence of a federal standard. A ruling is expected in early 2027. Combined with the federal copyright litigation and the pending DOJ records dispute, it will begin to define whether AI companies can continue to treat their training pipelines as proprietary information or whether courts and legislatures will require public accounting of what goes into the models that now mediate how hundreds of millions of people read, write, and find information.
The document request deadline at the end of November is the first checkpoint. The choice the Justice Department makes in the next ten days will tell a great deal about how the new administration plans to handle congressional oversight of its AI policy decisions, and about whether the September statement was an institutional act of government or a personal favor dressed in an official filing.
When a government body submits a legal position without consulting the agencies that wrote the law it is interpreting, the resulting document represents a political preference, not a legal analysis.
Key Takeaways
- 12 House Democrats sent a letter to AG Todd Blanche on November 19 demanding records on how the DOJ developed its September 1 fair use statement in the NYT v. OpenAI copyright case.
- Neither the Copyright Office nor the USPTO was consulted before the September filing, and both agencies' positions on AI training differ from the DOJ's broad fair use claim.
- No career antitrust attorneys signed the September DOJ filing, an unusual omission that legal observers say signals it bypassed the department's normal review process.
- Three House Judiciary Democrats filed a court amicus brief on October 30 calling the DOJ's position a "bulldozer" that disregards federal copyright law, before this week's larger letter with twelve signatories.
- Democrats take the House majority in January, converting this document request into a potential subpoena and extending the investigation into the new Congress if the department does not respond.
Questions Worth Asking
- If the court discounts the DOJ's September filing because of procedural questions, does that make OpenAI's fair use defense weaker than it would have been without any government intervention at all?
- The Trump administration's March 2026 AI policy framework said courts should decide the training data question. What changed between March and September, and who made that call inside the department?
- If Congress passes legislation requiring licensing fees for AI training data, would that apply retroactively to models already in commercial use, or only to future training runs?