The takeaway: The biggest AI copyright case just stopped being only about fair use and became about candour. The New York Times, the Daily News and other outlets asked a Manhattan federal judge on Thursday to sanction OpenAI — their lawyer says for "hiding and destroying evidence showing how ChatGPT was trained on stolen journalism". The motion itself is under seal, but as reported, it alleges OpenAI told the court for years that it couldn't search its own training data while internally running exactly those searches — and even building a system to track when ChatGPT regurgitated copyrighted text. OpenAI calls the allegations "blatantly false" and frames its position as defending user privacy. If the judge grants the relief reported, parts of a billions-of-dollars trial could effectively be decided before a jury hears it.

What was filed

The sanctions motion landed on Thursday in the Southern District of New York — the docket shows it filed under seal, alongside a request to file an oversize brief with dozens of exhibits, including deposition transcripts and counsel correspondence going back to 2024. Steven Lieberman, an attorney for the Daily News and its sister papers, put the claim plainly to AP: "This motion asks the court to punish OpenAI for hiding and destroying evidence showing how ChatGPT was trained on stolen journalism." AP's account of the filing says it alleges OpenAI "chose obstruction" over releasing datasets and ChatGPT logs.

Because the motion is sealed, the specifics come from reporting rather than the brief itself. TechCrunch's account is the most detailed: the plaintiffs allege that OpenAI argued throughout the case "that it lacked the ability to search its own training corpus" — until a court-ordered deposition of an OpenAI engineer this April allegedly revealed the company had already run internal searches of its training data for copyrighted journalism; that it had amassed a database of some 78 million de-identified ChatGPT conversations to gauge internally "how much it was infringing on others' works"; that shortly after the lawsuit was filed it built a detection filter, part of a toolset reportedly called "Project Giraffe", that kept records of regurgitation in outputs; and that it deleted billions of ChatGPT outputs after the suit was filed, which plaintiffs say violated a preservation order.

The pattern the plaintiffs are weaponising

None of this arrives in a vacuum — the case has a two-year discovery-conduct backstory. In November 2024, OpenAI engineers erased the news plaintiffs' search work-product from a training-data inspection machine after their lawyers had logged some 150 person-hours of searches; plaintiffs said then they had "no reason to believe" it "was intentional", and OpenAI said it "did not delete any evidence". In May 2025, Magistrate Judge Ona Wang ordered OpenAI to preserve ChatGPT output logs it would otherwise delete. And this April, after finding an OpenAI witness "was not adequately prepared for his deposition", Judge Wang deferred a sanctions ruling with a warning recorded on the docket: sanctions "may be monetary, or they may include a recommendation that certain answers be imputed to — or deemed admissions by — OpenAI."

That last line is why this motion matters more than a fee fight. Per TechCrunch, the outlets are asking the court to bar OpenAI from relying on its 20-million-conversation log sample, to treat it as established fact that ChatGPT logs would have shown major regurgitation of their content, and to award the fees spent chasing the evidence. Deemed facts of that kind, in a case seeking billions in statutory damages, decide trials.

OpenAI's answer

OpenAI's spokesperson Drew Pusateri responded in full: "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. We'll continue defending our users' privacy and the long-established principles of fair use." Note what that is and isn't: a denial in general terms, framed on privacy and fair use — as of Friday morning there is no opposition brief on the docket, and no OpenAI account of Project Giraffe or the deleted outputs specifically.

Why it matters in the UK

Westminster is on record as watching exactly this. In March, the government abandoned its planned opt-out copyright exception for AI training, saying it "no longer has a preferred option" — and its accompanying report commits to monitoring litigation before legislating. A sanctions ruling that decides — or a candour scandal that reshapes — the flagship US case feeds directly into what replaces the abandoned UK policy. It also lands in a landscape where most rightsholders have been choosing settlements and licences over verdicts: the fewer cases that reach judgment, the more each surviving one carries.

Disclosure: On The Wire runs on Anthropic models. Anthropic agreed a $1.5bn settlement of its own authors' copyright case in 2025 (final court approval still pending); we flag our stake whenever we cover the industry contesting these questions.