There is a word for what the New York Times accused OpenAI of on July 9, 2026. It is not a small word.
The word is lying. And the allegation is not vague. According to a 52-page filing in federal court in Manhattan, OpenAI told the court, for two years, repeatedly, that it could not search its training datasets or its ChatGPT output logs for copyrighted journalism. It said the task was technically burdensome. It said it raised privacy concerns. It said the tools did not exist.
Then a deposition happened. And OpenAI's own privacy engineer testified that the tools had existed all along. That OpenAI had already used them. That the searches had already been run. That a database of 78 million conversations, assembled specifically to measure OpenAI's own copyright exposure, had been sitting there, undisclosed, from before the lawsuit was even filed.
The publishers are not asking the court to rule in their favour on copyright. They are asking the court to punish OpenAI for what they call a deliberate and systemic effort to destroy and conceal the evidence.
Two Years of Litigation, One Deposition That Changed Everything
The New York Times filed its copyright lawsuit against OpenAI and Microsoft on December 27, 2023, the first news publisher to take legal action against a generative AI company. Assigned to Judge Sidney H. Stein in the Southern District of New York, it is the most watched copyright case in the AI industry.
The core claim: OpenAI trained ChatGPT on millions of Times articles without permission. ChatGPT then reproduced that journalism in its outputs, verbatim in some cases, directly competing with the Times as an information source and siphoning the web traffic that funds journalism. The Times was later joined by the Daily News, Center for Investigative Reporting, The Intercept, and Ziff Davis (CNET's parent), each filing separate suits with substantially similar allegations.
"This motion asks the court to punish OpenAI for hiding and destroying evidence showing how ChatGPT was trained on stolen journalism.", Steven Lieberman, attorney for the Daily News and seven sister papers. "OpenAI chose obstruction over just producing that evidence at the start of the case.", Plaintiffs' 52-page filing, SDNY, July 9, 2026.
Project Giraffe, 78 Million Conversations, and the Deposition That Blew the Cover
Throughout the case, OpenAI told the court a consistent story: searching its systems for copyrighted content was technically burdensome; producing ChatGPT logs raised user privacy concerns; the tools to do so did not exist. On this basis, the publishers' access to evidence was heavily restricted.
Then Vinnie Monaco, OpenAI's lead privacy engineer, testified under oath in a court-ordered deposition. According to the publishers' filing, Monaco revealed:
OpenAI had already conducted multiple internal searches of its training corpus for the plaintiffs' copyrighted journalism, searches it told the court it was unable to perform.
Before the NYT even filed its lawsuit, OpenAI had assembled a database of approximately 78 million de-identified ChatGPT conversations that it was using internally to assess how much it was reproducing others' copyrighted works.
After the lawsuit was filed, OpenAI implemented a Bloom filter as part of a toolset called Project Giraffe, a system designed to detect and record when ChatGPT was reproducing copyrighted content. Project Giraffe was never disclosed in discovery.
OpenAI had deleted billions of ChatGPT conversations, an alleged violation of the court's preservation orders requiring the company to retain relevant records.
The publishers had originally sought 120 million chat logs. OpenAI negotiated this down to 20 million. Then, rather than producing the 20 million it had agreed to, it offered only keyword-based search results instead, forcing three further court orders culminating in a January 5, 2026 ruling from Judge Stein.
The Full Landscape: Who Is Suing Whom
| Plaintiff(s) | Defendant | Filed | Status: July 2026 |
|---|---|---|---|
| New York Times | OpenAI + Microsoft | Dec 2023 | Discovery; sanctions motion filed July 9, 2026 |
| Daily News, Ziff Davis, CIR, The Intercept | OpenAI | 2024 | Consolidated with Times case; party to July 9 motion |
| New York Times | Perplexity | 2025 | Active; separate from OpenAI case |
| Bartz et al. (~500K authors) | Anthropic | 2023 | $1.5B class settlement announced Sep 2025; awaiting final approval |
| Authors Guild + individual authors | OpenAI | 2023 | Active; discovery phase |
| Visual artists | Stability AI, Midjourney, DeviantArt | 2023 | Active; multiple courts |
| UMG, Sony, Warner | Suno, Udio | 2024 | Active; music AI copyright test cases |
The Bartz v. Anthropic settlement is the most significant data point in the entire landscape. Anthropic downloaded approximately 500,000 books from pirate libraries to train its models. After winning on fair use at summary judgment, it agreed in September 2025 to a $1.5 billion class settlement, approximately $3,000 per covered work, requiring it to destroy the pirated files and granting no licence for future training use. Final approval is pending. This is the largest AI copyright payout to date and provides an informal benchmark for every negotiation in the field.
Why This Case Is Bigger Than the Times
The Times has spent more than $28 million fighting AI companies, $4.2 million in the first quarter of 2026 alone. That figure is not just a legal budget line. It is a statement about existential stakes.
The threat is structural. When Google introduced AI-generated summaries in 2024, publishers lost traffic. When ChatGPT reproduces an article, users do not click through to the original. The journalism still exists. The revenue model that funds it does not.
The industry is negotiating and suing simultaneously. The Times struck a multiyear licensing deal with Amazon in May 2025, reportedly worth $20 to $25 million for AI uses of its content. Other publishers have signed with OpenAI, Google, and Meta. The same companies are defendants in Manhattan federal court and licensing partners in the same building on different floors.
The White House's March 2026 National Policy Framework took OpenAI's side on the central legal question, stating AI training does not violate copyright law and preferring court resolution over mandatory licensing. That political signal makes a merits win for publishers harder. But the July 9 sanctions motion is not about the merits. It is about whether OpenAI corrupted the process by which the merits could be determined.
If the court grants even part of what the publishers are asking, accepting as established fact that ChatGPT logs would show major regurgitation; barring OpenAI from using its own log sample as defence evidence, the shape of the case changes permanently. Sanctions are not just punishment. In civil litigation, they become facts. And facts established by a court are very hard to undo on appeal.
After the AI Hype
The bubble, the self-dealing, and what AI actually delivers versus what it promises. The article that started AI Angst.
NYT vs OpenAI: FAQ
The New York Times filed a copyright infringement lawsuit against OpenAI and Microsoft on December 27, 2023, in the US District Court for the Southern District of New York (Judge Sidney H. Stein). It is widely considered the most consequential test of whether training AI on copyrighted journalism constitutes fair use under US copyright law. The Times was joined by the New York Daily News, the Center for Investigative Reporting, The Intercept, and Ziff Davis (CNET's parent), each filing separate suits with similar allegations.
In a 52-page filing, publishers allege a "deliberate and systemic effort to obstruct discovery": (1) OpenAI falsely told the court for two years it lacked tools to search training data and output logs for copyrighted content; (2) OpenAI secretly possessed a database of approximately 78 million de-identified conversations used internally to measure copyright exposure; (3) OpenAI implemented a Bloom filter (Project Giraffe) to detect content regurgitation, never disclosed in discovery; (4) OpenAI deleted billions of relevant conversations in violation of court preservation orders.
Vinnie Monaco is OpenAI's lead privacy engineer. In a court-ordered deposition, he allegedly revealed that OpenAI had conducted multiple internal searches of its training corpus for copyrighted journalism, despite telling the court it could not do so, and that OpenAI had assembled a database of approximately 78 million de-identified ChatGPT conversations to measure its own copyright exposure before the NYT lawsuit was even filed. Monaco's deposition is the evidentiary foundation of the July 9, 2026 sanctions motion.
Project Giraffe is an internal OpenAI toolset that included a Bloom filter, a probabilistic data structure used to detect when ChatGPT outputs matched or closely reproduced training data content. The publishers allege OpenAI implemented it shortly after the NYT filed its lawsuit in December 2023 to track content regurgitation in ChatGPT outputs. Project Giraffe was allegedly never disclosed during discovery, despite being directly relevant to whether ChatGPT reproduces copyrighted journalism.
Four remedies: (1) Prevent OpenAI from using its 20-million chat log sample as evidence, on grounds of alleged unreliability due to deletions; (2) Accept as established fact that ChatGPT logs would have shown major regurgitation of copyrighted content; (3) Prevent OpenAI from arguing its logs do not demonstrate substantial reproduction; (4) Require OpenAI to pay plaintiffs' attorneys' fees incurred in pursuing concealed evidence. Publishers also request any additional deterrence penalties the court deems appropriate.
OpenAI spokesperson Drew Pusateri denied all allegations: "As the Times' case weakens and they've been forced to drop claims against us, they're persisting with their efforts to invade the privacy of people who have nothing to do with this case, including by making these blatantly false allegations." OpenAI has consistently argued that producing chat logs risks user privacy violations, that searching training data is technically burdensome, and that the Times has dropped some original claims. OpenAI has not publicly addressed the specific Project Giraffe or 78-million-conversation allegations.
More than $28 million in total litigation costs against AI companies, including $4.2 million in Q1 2026 alone. The Times has filed a separate lawsuit against Perplexity in addition to its OpenAI/Microsoft case. Despite the litigation, it struck a multiyear content-licensing deal with Amazon in May 2025 reportedly worth $20 to $25 million for AI uses, illustrating the paradox of simultaneously suing and licensing to the same AI ecosystem.
News publishers suing OpenAI alongside the Times: Daily News, Center for Investigative Reporting, The Intercept, and Ziff Davis (CNET parent). Beyond news: authors filed the Bartz class action against Anthropic; visual artists sued Stability AI, Midjourney, and DeviantArt; major music labels (Universal, Sony, Warner) sued AI music generators Suno and Udio. The Bartz v. Anthropic $1.5 billion class settlement (September 2025) covering approximately 500,000 books, roughly $3,000 per work, is the largest AI copyright payout and the benchmark for all subsequent settlement negotiations.
Whether training AI on copyrighted material constitutes fair use under US copyright law. OpenAI argues ingesting text to train a statistical model is transformative and does not substitute for the original in the market. Publishers argue ChatGPT directly competes with their journalism by reproducing it in outputs and reducing clicks to original sources. The White House's March 2026 National Policy Framework stated AI training does not violate copyright law, supporting court resolution rather than mandatory licensing. No trial date has been set; the case remains in discovery as of July 2026.
Bartz v. Anthropic was a class action by approximately 500,000 authors whose books Anthropic downloaded from pirate libraries to train its models. After winning on fair use at summary judgment, Anthropic agreed in September 2025 to a $1.5 billion class settlement, approximately $3,000 per covered work, requiring it to destroy the pirated files, granting no licence for future training use. As of mid-2026, it awaits final court approval. It is the largest AI copyright payout to date and establishes an informal pricing benchmark: every AI company assessing its copyright litigation exposure now does so with this number in mind.
Jans Bock-Schroeder
Publisher & Founder of AI Angst
Coming from the world of art, photography, and the luxury market, Jans launched AI Angst in 2025 to explore the cultural, ethical, and psychological impacts of artificial intelligence. His work bridges creative vision with critical technology analysis, offering clarity in an era of rapid technological change.
Sources and Citations
All sources published July 9–10, 2026:
-
TechCrunch: "New York Times says OpenAI hid evidence in ChatGPT copyright trial" (July 9, 2026)
Primary source for Project Giraffe, Bloom filter detail, 78M conversation database, Vinnie Monaco deposition content, and the four specific sanctions requested.
https://techcrunch.com/2026/07/09/new-york-times-says-openai-hid-evidence-in-chatgpt-copyright-trial/ -
Reuters / US News: "New York Times-Led Group Asks Court to Sanction OpenAI" (July 9, 2026)
Source for "falsely told the court," "even before the first News Plaintiff filed suit," "deleted billions of relevant ChatGPT conversations," and Drew Pusateri OpenAI denial quote.
https://money.usnews.com/investing/news/articles/2026-07-09/new-york-times-led-group-asks-court-to-sanction-openai-in-us-copyright-dispute -
Al Jazeera: "NYT-led group asks court to sanction OpenAI in US copyright dispute" (July 9, 2026)
Source for Steven Lieberman attorney quote, $28M total litigation cost figure, $4.2M Q1 2026 litigation cost, AI summary / traffic displacement context, and 52-page filing description.
https://www.aljazeera.com/economy/2026/7/9/nyt-led-group-asks-court-to-sanction-openai-in-us-copyright-dispute -
Variety: "New York Times and Other News Outlets Accuse OpenAI of Lying in Discovery, Demand Legal Sanctions" (July 9, 2026)
Source for full plaintiff list (Times, Daily News, CIR, The Intercept, Ziff Davis), "deliberate and systemic effort to obstruct discovery" quote, and "chosen obstruction" phrasing.
https://variety.com/2026/digital/news/new-york-times-news-outlets-accuse-openai-of-lying-lawsuit-1236805648/ -
Washington Post: "News outlets urge a judge to sanction OpenAI in a high-stakes AI copyright fight" (July 9, 2026)
Source for broader industry context: AI-generated summaries cutting publisher ad revenue, publishers walking line between litigation and licensing deals.
https://www.washingtonpost.com/business/2026/07/09/openai-new-york-times-ai-copyright-lawsuit/ -
LegalClarity: "New York Times vs. OpenAI Lawsuit Status and Timeline" (June 2026)
Source for case timeline: December 27, 2023 filing date, Judge Stein assignment, January 5, 2026 ruling, Bartz v. Anthropic $1.5B settlement detail, Amazon licensing deal value ($20–$25M), and White House March 2026 National Policy Framework on AI copyright.
https://legalclarity.org/new-york-times-vs-openai-lawsuit-status-and-timeline/
Published: July 12, 2026. Sources verified at time of publication. Case is ongoing, no sanctions ruling has been issued as of publication date. All external links open in a new tab.
