NYT Seeks Sanctions on OpenAI Over Hidden Evidence
The New York Times and Daily News asked a federal judge to sanction OpenAI for concealing evidence that its systems detect and log copyrighted content regurgitation.
6 min read
The New York Times and Daily News filed a motion on July 9, 2026 asking a federal judge to sanction OpenAI for what they call "discovery misconduct" - specifically, hiding evidence that the company had already run internal searches of its training corpus for copyrighted journalism and built tools to track when ChatGPT reproduced those works verbatim.
The motion, filed in Manhattan federal court, comes after an April deposition of Vinnie Monaco, an OpenAI data privacy engineer, produced a set of revelations that the publishers say contradict two years of statements from OpenAI's legal team.
"If OpenAI genuinely believed that copying our clients' journalism was fair and legal, it wouldn't have hid the truth about having done it."
- Ian B. Crosby, lead counsel for the plaintiffs
What the Deposition Revealed
The 78 Million Chat Logs
OpenAI had assembled a database of roughly 78 million de-identified ChatGPT conversations, which it used internally to assess how much its model reproduced copyrighted content. Publishers requested access to this data as part of discovery.
For over two years, OpenAI's legal team claimed it was technically impossible to search its training corpus or output logs for copyrighted material. Monaco's deposition contradicts that directly. According to the motion, OpenAI had already conducted exactly those searches - and held the results.
The original discovery request asked for 120 million chat logs. OpenAI negotiated that figure down to 20 million, then submitted a sample in December 2025 that the court described as so heavily redacted as to be "unusable." Publishers now allege that OpenAI also deleted billions of outputs during the litigation period and substituted millions of logs within the sample it eventually provided.
Project Giraffe and the Bloom Filter
Shortly after the Times filed its initial lawsuit in late 2023, OpenAI built a set of internal detection tools it called "Project Giraffe." The key component was a Bloom filter - a data structure used to check whether an output matches a set of known strings - that detected and logged instances where ChatGPT reproduced content from its training data.
Put differently, OpenAI was tracking regurgitation internally while telling the court it couldn't measure or identify it. The publishers argue that Project Giraffe was itself an admission that the company knew wholesale reproduction was happening.
The sanctions motion turns on whether OpenAI deliberately withheld evidence that would have shown widespread content reproduction.
Source: pexels.com
What Publishers Are Demanding
The Four Sanctions Sought
The Times and Daily News are asking the court for four specific remedies:
- Exclude the 20 million chat log sample as unreliable evidence, on the grounds it was redacted beyond usefulness and may have been curated to minimize apparent infringement
- Accept as established fact that a full review of ChatGPT logs would have shown major reproduction of the plaintiffs' journalism
- Bar OpenAI from arguing that the sample it provided doesn't demonstrate significant regurgitation
- Require OpenAI to pay the plaintiffs' legal fees incurred during the discovery dispute
The motion was joined by the Daily News's parent company, which filed a related lawsuit in April 2024 alongside a group of other newspaper publishers.
The Discovery Timeline
December 2023 - New York Times files suit against OpenAI and Microsoft in Manhattan federal court.
April 2024 - Daily News LP and other New York newspaper publishers file separate but related suit.
May 2025 - Court orders OpenAI to preserve and segregate all output log data that would otherwise be deleted.
December 2025 - OpenAI submits 20 million chat logs to the court. Plaintiffs call the redactions so extensive the sample is useless.
April 2026 - Deposition of Vinnie Monaco reveals the 78 million conversation database and Project Giraffe.
July 9, 2026 - Publishers file formal sanctions motion in Manhattan federal court.
Who Pays If the Motion Succeeds
Impact Assessment
| Stakeholder | Impact | Timeline |
|---|---|---|
| OpenAI | Key evidence arguments blocked; legal fee payments | August-September 2026 |
| Publishers (NYT, Daily News) | Stronger procedural position; regurgitation assumed proven | Immediate |
| Other AI companies | Sets discovery precedent for all copyright cases | Long-term |
| ChatGPT users | Privacy arguments weakened if logs become central evidence | Ongoing |
OpenAI's Defense
OpenAI spokesperson Drew Pusateri pushed back on the motion in a statement, framing the dispute as a privacy matter rather than a transparency failure.
"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."
- Drew Pusateri, OpenAI spokesperson
The company's broader defense rests on two arguments: that user privacy prevents sharing chat logs in detail, and that training on publicly available journalism falls under fair use. Pusateri's statement invoked both.
The fair use argument is the central question for the trial. But the sanctions motion is a separate, procedural fight - one that could notably reshape how that evidence is presented before a jury ever hears the case.
This case sits with several others pressing similar claims. Britannica and Merriam-Webster sued OpenAI earlier this year, alleging ChatGPT copied nearly 100,000 articles and dictionary entries without a license. Separately, researchers have shown that frontier models can reproduce copyrighted books from memory at surprisingly low cost - evidence that could become relevant across multiple pending suits.
OpenAI's fair-use defense is the central question at trial, but the sanctions fight is a separate procedural battle that could reshape how evidence is presented.
Source: pexels.com
What Happens Next
The judge must rule on the sanctions motion before the case proceeds to trial. A favorable ruling for the publishers wouldn't end the lawsuit, but it would shift the evidentiary ground markedly - OpenAI would be blocked from using its own (disputed) chat log sample to argue minimal regurgitation, and the court would treat major reproduction as an established fact.
Trial is currently expected in late 2026 or early 2027. If the sanctions are granted, OpenAI enters that trial having already conceded the core quantitative question the publishers are trying to prove.
The outcome matters beyond this case. Every major AI company trained on web-scraped content is watching how courts handle the discovery obligations that come with that training data. A ruling that penalizes OpenAI for concealing detection tools sets a standard for what future defendants are expected to disclose - and what they can claim they can't search.
Sources:
