Where does the NYT v. OpenAI lawsuit stand now?

Status as last verified in late August 2026

The status starts with the docket caption and posture: The New York Times Company v. Microsoft Corporation et al., No. 1:23-cv-11195, pending in the Southern District of New York before District Judge Sidney H. Stein and Magistrate Judge Ona T. Wang, filed Dec. 27, 2023. The docket reflects a live case, not a final copyright ruling; the complaint includes a jury demand and damages requests that remain allegations and requested relief, not adjudicated amounts.[1]

As of late August 2026, the core infringement claims had survived the motion to dismiss; major discovery orders requiring preservation and production of ChatGPT output-log material had been affirmed; a proposed Microsoft-related amendment had been narrowed by order; the publishers’ sanctions motion and OpenAI’s opposition were on file; summary-judgment briefing was reported complete; and no trial date had been set.[2][3][4][5][6][7]

What the court has actually decided

The motion-to-dismiss ruling is the most commonly overstated event in the case. It did not decide whether OpenAI or Microsoft infringed copyrights. It did decide that the Times’ direct and contributory copyright-infringement theories could proceed past the pleading stage, while trimming other claims. In the opinion dated Mar. 26, 2025 and filed Apr. 4, 2025, the court allowed the direct and contributory infringement claims to survive, dismissed the DMCA § 1202 claims with leave to amend, dismissed common-law unfair competition with prejudice, and rejected the defendants’ three-year statute-of-limitations argument as to the challenged 2019–2020 training conduct at that stage.[2]

Those verbs matter. “Survived” means the claims remained legally sufficient enough to move into discovery and later merits testing. It does not mean the Times proved copying, market harm, lack of fair use, contributory liability, or damages. “Dismissed with leave” means the plaintiff may try to replead. “Dismissed with prejudice” means that claim is out unless a higher court later changes the result. That is the usable status for briefs, procurement memos, and partner updates.

The discovery rulings have more immediate operational force. A May 13, 2025 preservation order required OpenAI to preserve output logs that were otherwise slated for deletion, in a service environment described in reporting as involving more than 400 million users; that order was affirmed on Jun. 26, 2025. Later, a Nov. 7, 2025 order compelled production of 20 million de-identified ChatGPT logs, and Judge Stein affirmed that order on Jan. 5, 2026 as “neither clearly erroneous nor contrary to law.”[3]

For litigation-risk purposes, the log rulings are not a copyright merits holding. They are, however, concrete discovery consequences. They show that AI output logs can become preservation and production targets when they are tied to infringement theories and user-output evidence. That makes this case relevant to AI logs as discoverable ESI and to litigation-hold workflows for AI-generated documents, even though it has not yet supplied the final copyright rule that many summaries want it to supply.

What remains pending

The sanctions motion is pending, not decided. On July 9, 2026, a Times-led publisher group filed a memorandum seeking sanctions against OpenAI. The moving papers name, among others, the Daily News, the Center for Investigative Reporting, The Intercept, and Ziff Davis, and accuse OpenAI of discovery misconduct involving searchability, training-data representations, and log evidence. Those are the movants’ allegations and requested consequences; they are not court findings.[4]

OpenAI’s Aug. 14, 2026 opposition was reported as arguing that the Times sought sanctions because it lacked evidence. That filing, too, is advocacy unless and until the court adopts part of it. The docket significance is that the sanctions issue had been fully joined enough for the court to address, not that either side’s account had been accepted.[6]

Summary judgment was also unresolved in the verified late-August-2026 posture. A self-published AI litigation tracker reported that summary-judgment briefing concluded around Apr. 2, 2026 and projected a possible late-2026 or 2027 trial window if claims survived. That is useful as a tracking signal, but it should not be cited as a scheduling order. The docket source used for this status did not supply a verified trial date.[1][7]

That leaves two pending rulings with practical force: sanctions and summary judgment. Sanctions could affect evidentiary presumptions, case management, or other litigation consequences if granted. Summary judgment could define which copyright theories, defenses, and parties reach trial. Until those orders issue, the case remains past the headline-heavy pleading and discovery fights but short of the ruling that would fix its merits meaning.

The Microsoft amendment should be treated as a narrowing event, not a verdict

The Aug. 6, 2026 order at Dkt. 1449 is another place where shorthand can do damage. The order denied in part the Times’ June 2026 attempt to add a Microsoft “supercomputer” contributory-infringement theory associated with the Supreme Court’s Cox standard. The usable status is that the proposed amendment was narrowed; the order should not be recast as a merits ruling that Microsoft’s infrastructure either did or did not create copyright liability.[5]

Anyone quoting this point in a filing should open Dkt. 1449 and quote the operative language rather than rely on secondary headlines. The important distinction is between pleading permission and liability. A denied amendment may limit what theory goes forward; it does not, by itself, adjudicate the claims that remain in the case.

How to use this case without overstating it

The case is strong support for a narrow proposition: publisher claims against AI developers and infrastructure partners can survive early dismissal when pled with enough specificity, and AI-output log evidence can become a serious preservation and production issue. It is not yet authority for the broader proposition that training on news content is or is not fair use, that model outputs establish infringement at scale, or that any particular damages number is recoverable.

For adjacent AI-training copyright risk, this record sits beside other publisher and music-rights disputes rather than replacing them. The better comparison set includes the GEMA v. Suno Risk Digest record. For discovery and confidentiality questions, it should be read alongside the site’s AI eDiscovery privilege rulings record and the Sam Altman ChatGPT privilege briefing, because the recurring operational question is not only copyright exposure but who preserved what, who searched what, and which AI-system records the court treats as evidence.

Party pages, press statements, and self-published trackers can help orient the timeline. They should not be allowed to outrank orders, docket entries, or filed briefs when describing what has been fixed by court action. The late-August-2026 posture is therefore limited but important: core claims survived, discovery orders landed, the Microsoft amendment was narrowed, sanctions and summary judgment remained pending, and no trial date was set.

References

1. The New York Times Company v. Microsoft Corporation, CourtListener.
2. The New York Times Company v. Microsoft Corporation et al, Dkt. 514, Justia.
3. ChatGPT creator must turn over 20M chat logs in copyright litigation, federal judge says, ABA Journal, Jan. 8, 2026.
4. Memorandum of Law in Support of Motion for Sanctions Against OpenAI, July 9, 2026.
5. Order, Dkt. 1449, Aug. 6, 2026.
6. OpenAI Says NY Times Seeks Sanctions Due to Lack of Evidence, Bloomberg Law, Aug. 14, 2026.
7. New York Times v. OpenAI, AI Lawsuit Tracker.