Invest1 distinct publisher3 min readPublished
The Statement of Interest rests its fair use claim on two executive orders and a statute that requires no leave to file, which tells you what the filing cost the government and how long any relief priced off it can last.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
A brief that needs neither leave nor a deadline is cheap to file, and the document says so itself, citing Gill v. Winn Dixie Stores for the proposition that section 517 contains no time limitation and does not require the court's leave [5]. That low cost still bounds what the paper can be: the United States did not become a party, and in footnote 2 it declined to contend that anything alleged in the litigation was authorized or consented to by the government, or done for the benefit of the United States, as set forth by 28 U.S.C. section 1498 [4]. The government stopped short of intervening or of claiming the training as its own procurement [4].
The interest claim is broad where the disclaimer is narrow: a strong interest in whether training AI models on written works constitutes fair use under copyright law [3], plus a stated interest in an AI industry that sets the standard for the practice and procedure of AI use globally [10], resting on the January 2025 order's language about sustaining and enhancing America's global AI dominance [7] and a June 2, 2026 order about encouraging AI innovation and accelerating responsible AI adoption across government and industry [8]. Ninety-one days separate that second order from the filing [2]; the first is nineteen months and nine days back [3]. The government's paper arrived as document 1,682 [1], which is 1,681 filings into one consolidated proceeding [4], and it runs 20 pages of which the published excerpt stops at page 2 [6], leaving 18 pages of actual argument outside this record [1].
For anyone marking a reserve against a model developer, the number that moves is expected damages, and an executive branch can declare an interest without supplying deference on a fair use question. The collateral here is two executive orders [7][8], revocable by the office that issued them, so a litigation discount traced to this filing carries the duration of the policy rather than of the copyright term.
If the court quotes the statement in its fair use holding, then 20 pages [6] bought more than 20 pages usually buys. If instead the court treats it as the views of a non-party, the real audience becomes Congress and the settlement table, where counsel across all matters in the MDL [1] mark trial value down by whatever they judge an executive preference to be worth. Plaintiffs could also turn footnote 2 around, arguing that a government unwilling to claim the training as authorized or for its benefit [4] took the cheap half of its own position.
The view these two pages support is a narrow one: the filing addresses the political tail, a statutory carve-out or an agency move against training, and leaves the modal outcome, a damages figure set by a court, close to where it sat on August 31. An order that cites the Statement of Interest as a reason for its holding would break that read, and so would a settlement struck on terms showing plaintiffs cut their number after September 1 [1].
Ranked by verification strength, evidence, and original report placement.
The filing states that "The United States thus has a strong interest in the question whether training AI models on written works constitutes 'fair use' under copyright law."
The filing's page headers read "Page 1 of 20" and "Page 2 of 20", and the published excerpt covers only those first two pages.
The Intercept published the document under the headline "Intercept Copyright Lawsuit: Trump Department of Justice Briefing in Support of OpenAI".
The filing states that 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.
Document 1682, a Statement of Interest of the United States, was filed on 09/01/26 in the U.S. District Court for the Southern District of New York in Case 1:25-md-03143-SHS-OTW, captioned as copyright infringement litigation, and states "This Document Relates To: All Matters".
The document states: "The United States appears pursuant to 28 U.S.C. § 517 to inform the Court of its interest in this litigation."
Distinct publishers with included, body-backed reporting in this cluster.
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Federal government asks SDNY to treat LLM training alone as transformative fair use1 distinct publisher
invest
The Justice Department moves the AI copyright fight to the licensing table1 distinct publisher
invest
Storm's reductio: if the DOJ theory catches him, it catches OpenAI and Google1 distinct publisher
build
DOJ asks the court to judge LLM training separately from model outputs2 distinct publishers
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Stamped and filed, but roughly scanned and mostly unpublished
The best feature of this story is that it rests on the document rather than a description of one: case number, document number, filing date, the § 517 footnote and both executive-order citations can be read directly. The limits are equally concrete. The transcription renders the caption as "PENAL, INC." and Gill v. Winn-Dixie as "Gil: Winn Dixie Sores", and it stops mid-sentence on page 3 of 20. Quote-level accuracy is adequate for the interest section and simply unavailable for everything the brief does after it.
Nothing to count three days in
A brief filed on September 1 and published on September 4 has no takeup to measure. Our coverage shows no ruling on it and no response from any party, and the only thing resembling a measure of uptake would be what the court does with the government's view.
"In support of OpenAI" is the headline's phrase, not the brief's
The Intercept titles this a briefing in support of OpenAI. The reproduced pages say the United States has a strong interest in the fair-use question and then move to national security; the department's answer, if it gives one, is in the seventeen pages that were not posted. Our own framing inherits part of that gap. The executive-order footing and the no-leave statute are visible and quotable; the conclusion that the filing therefore backs one party is inference sitting on a headline.
Filer and publisher each have a side
The department's stated interest is the policy of the president it serves, quoted twice from executive orders, and § 517 meant it needed nobody's permission to file, so the brief carried little procedural cost. On the other side of the page, the outlet publishing it calls the case the Intercept copyright lawsuit, which makes this a plaintiff circulating a document that argues against its own position. Neither fact makes the quoted words less accurate, but both explain what was selected and how it was titled.
Firm on the words, thin on the argument
What the first pages say, and which statute the government invoked, is close to certain. What the brief concludes, whether it engages the plaintiffs' claims at all, and how the court received it are all outside what we have, from a single publisher's scan. That certainty about the opening pages does not extend to the argument they introduce.