The New York Times just made its biggest legal move yet against OpenAI. This week, the Times and more than a dozen other newspapers asked a federal judge in Manhattan to sanction OpenAI, accusing the company of hiding and destroying evidence about how ChatGPT was trained on their journalism. It’s the sharpest escalation yet in the New York Times OpenAI lawsuit, a legal fight that’s dragged through discovery since 2023 and just got a lot more personal in 2026.
Here’s the honest truth about why this matters: lawsuits like this usually move in slow, boring increments — motions, depositions, redacted filings. This one didn’t. The publishers are accusing OpenAI of lying to the court, and they say they can prove it.
What the Publishers Say OpenAI Hid
According to the motion, filed Thursday, July 9, 2026, in the U.S. District Court for the Southern District of New York, and first reported by TechCrunch, the case centers on an April deposition of Vinnie Monaco, an OpenAI data privacy engineer. Monaco reportedly testified that OpenAI had already searched and evaluated its training data for the plaintiffs’ copyrighted journalism — and had done so well before the Times ever filed suit. The publishers say that directly contradicts OpenAI’s earlier claims to the court that it lacked the technical ability to run those searches.
The filing also alleges OpenAI had quietly amassed a database of roughly 78 million de-identified ChatGPT conversation logs, used internally to gauge how much of its output echoed other people’s copyrighted work. On top of that, the publishers claim OpenAI built a detection system — a “Bloom filter,” part of a broader internal effort reportedly called “Project Giraffe” — specifically to flag and log regurgitated content in ChatGPT’s responses, and that it did so shortly after the lawsuit was filed. The motion goes further, alleging OpenAI deleted logs in violation of the court’s evidence-preservation order.
What Sanctions They’re Actually Asking For
This isn’t just a “please behave” filing. According to court documents, the publishers want the judge to: bar OpenAI from using its 20-million-chat sample as evidence, since they argue it’s unreliable; treat it as an established fact that the full logs would show substantial regurgitation of their content; stop OpenAI from arguing the sample doesn’t show meaningful copying; and make OpenAI cover the legal fees the publishers racked up chasing this evidence down. That’s a mix of evidentiary sanctions and a monetary penalty — the legal equivalent of the judge assuming the worst about OpenAI’s conduct before the case even reaches trial.
How the New York Times v. OpenAI Lawsuit Got Here
If you haven’t followed this case since the beginning: the Times sued OpenAI and Microsoft back in December 2023, arguing the companies trained their AI models on millions of Times articles without permission or payment, and that ChatGPT and Bing Chat could reproduce near-verbatim chunks of paywalled reporting. OpenAI’s defense has leaned entirely on fair use, and in 2025 it partly worked — a judge narrowed some claims — but the core copyright allegations survived and are still headed toward resolution. In January 2026, the presiding judge upheld an order compelling OpenAI to hand over 20 million anonymized ChatGPT logs to the plaintiffs, which is the same dataset now at the center of this sanctions fight.
OpenAI’s business has kept moving fast regardless of the lawsuit hanging over it. It launched an autonomous work agent this year — we broke down what that means for you in our coverage of the ChatGPT Work launch. It followed that up with a real-time voice assistant, which we covered in detail in our GPT-Live launch piece. None of that product momentum makes this legal exposure disappear.
OpenAI’s Response
OpenAI is pushing back hard. “As the Times’ case weakens and they’ve been forced to drop claims against us, they’re persisting with efforts to invade the privacy of people” unrelated to the case, spokesperson Drew Pusateri said, calling the allegations “blatantly false.” OpenAI has previously said it will keep defending “our users’ privacy and the long-established principles of fair use.”
What’s Next
The judge hasn’t ruled on the sanctions motion yet, and there’s no confirmed trial date for the underlying case. What happens next matters beyond this one lawsuit: it’s part of a much bigger legal reckoning for AI training data. Authors already settled a similar fight with Anthropic for $1.5 billion, Getty just struck a licensing deal with OpenAI instead of continuing to fight it in the U.S., and the Copyright Alliance counts more than 70 active AI copyright suits industry-wide. If the judge grants even part of what the publishers are asking for, it could hand the Times a serious advantage heading into trial — and give every other publisher suing an AI company a new playbook.
It also arrives during a stretch when OpenAI is already facing outside scrutiny on other fronts. The 2026 AI Safety Index, released this week, gave the company one of the lowest grades of any major lab. We’ll keep tracking this case as it moves toward whatever comes next — sanctions ruling, settlement, or trial.
Deixe um comentário