The New York Times, New York Daily News, and 15 other media organizations filed a motion Thursday in Manhattan federal court seeking sanctions against OpenAI, alleging the company hid its ability to search ChatGPT for copyright violations for more than two years. The motion cites an April 2026 deposition of OpenAI data privacy engineer Vinnie Monaco, who revealed the company had built internal tools to detect when its outputs reproduced copyrighted content. OpenAI had also amassed 78 million internal ChatGPT conversations to measure how much it was infringing — while telling the court such searches were technically impossible. The Times filed the original lawsuit in December 2023. The court hasn't set a trial date.

1. OpenAI Knew What It Had (Ian B. Crosby, Susman Godfrey; Steven Lieberman, Rothwell Figg)

Publishers say the Monaco deposition turned a copyright case into a perjury case.

OpenAI built the detection tools and then said it didn't have them. The motion's core allegation is that OpenAI falsely claimed for more than two years it couldn't search its systems for copyrighted journalism — while internally running exactly those searches. "For over two years, OpenAI lied to the Times, the Daily News plaintiffs, the public and the court," said Ian B. Crosby, lead counsel for the Times at Susman Godfrey.

Project Giraffe is the smoking gun. After publishers sued, OpenAI built "Project Giraffe" — a set of tools including a "Bloom" filter to detect when ChatGPT was reproducing copyrighted content. Publishers say this proves OpenAI wasn't technically limited from the start; it was strategically withholding.

OpenAI also destroyed evidence. The company allegedly deleted billions of ChatGPT outputs after the suit landed, then sent a 20 million log sample with so many redactions that the court called it "unusable." Rothwell Figg attorney Steven Lieberman said the motion "asks the court to punish OpenAI for hiding and destroying evidence showing how ChatGPT was trained on stolen journalism." Publishers want the court to declare, as established fact, that the logs would have shown substantial reproduction — removing OpenAI's ability to argue otherwise at trial.

2. But OpenAI Says This Is a Privacy Fight (Drew Pusateri, OpenAI)

OpenAI says publishers are trying to dig through private user conversations as their case runs out of steam.

OpenAI says publishers are after private user data, not evidence. Spokesperson Drew Pusateri said publishers are "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 notes the Times has been "forced to drop claims" as the litigation has progressed.

OpenAI says the redactions protected users, not evidence. District Judge Sidney Stein rejected that argument in January 2026 — he ordered all 20 million logs produced, finding users "voluntarily submitted" their conversations to OpenAI. But OpenAI maintains the privacy concern is genuine and that producing more chat data would harm real users.

Fair use is still the core argument. OpenAI says training a large language model on publicly available text is transformative use under copyright law, regardless of what the output logs show. Courts are developing a consensus that general-purpose AI training is "highly transformative" — a factor favoring fair use.

3. The Rest of the AI Industry Is Watching (Joseph C. Gratz, Morrison & Foerster; AI Lawsuit Tracker)

The sanctions motion could short-circuit the fair use question before it ever reaches a jury.

If the court grants sanctions, the judge declares infringement proven. Publishers are asking the judge to establish as a matter of law that OpenAI's logs would have shown substantial copyright reproduction. That would strip OpenAI's fair use defense of its most important piece — the argument that the model doesn't substantially substitute for the original works.

The remedies the Times is seeking would reshape the industry. The Times wants billions in statutory damages — up to $150,000 per work under willful infringement rules — and is asking the court to order OpenAI to destroy the GPT models trained on its content entirely. No court has ever ordered an AI company to destroy its models.

Google, Meta, and Anthropic all face copyright suits too. Morrison & Foerster partner Joseph C. Gratz notes that "sharp disagreements persist" among courts on how to handle AI training, and that 2026 is unlikely to bring final answers to the copyright questions. The outcome of the evidence fight here will shape what any of those companies has to hand over in discovery.

Where This Lands

Publishers have made the evidence fight the first thing the judge will rule on. They say Monaco's deposition proves OpenAI knew what its systems were doing and chose not to tell the court. OpenAI says this is a weakening lawsuit dressed up as a misconduct case, and that user privacy is a real concern the courts have already had to weigh. However the judge rules, it sets the floor for what Google, Meta, and Anthropic have to hand over in their own cases.

Sources