The New York Times is urging a federal judge to sanction OpenAI for allegedly hiding ChatGPT logs relevant to a copyright lawsuit. The news outlets claim OpenAI is not complying with discovery requests in a landmark case about AI training data.
SynthePulse Insight · AI deep reading
Version 1 · 3 sources
The New York Times and Daily News accuse OpenAI of systematically hiding evidence in a copyright lawsuit, including an internal tool called 'Project Giraffe' and 78 million de-identified chat records. Plaintiffs ask the court to sanction OpenAI and presume its content was heavily copied.
On July 9, 2026, The New York Times and Daily News filed a motion in court accusing OpenAI of systematically hiding evidence during a two-year copyright lawsuit. Plaintiffs claim that OpenAI not only falsely stated it could not search its training corpus but also secretly developed a toolset called 'Project Giraffe,' which includes a 'Bloom' filter used to detect and record instances of 'regurgitation' in ChatGPT outputs. The tool was deployed shortly after the lawsuit was filed, but OpenAI never disclosed it to plaintiffs or the court.
More critically, OpenAI data privacy engineer Vinnie Monaco testified in a court-ordered deposition in April that OpenAI had established a database containing approximately 78 million de-identified ChatGPT conversations before the Times filed its lawsuit, used for internal evaluation of how much the model infringed on others' works. Plaintiffs had previously requested a sample of 120 million chat records, but OpenAI, citing technical burden and user privacy, eventually reduced the sample to 20 million records, which it submitted in December 2025. However, plaintiffs claim the sample was heavily redacted and overly blacked out, and the court found it 'unusable.'
Plaintiffs' attorney Ian B. Crosby said: 'If OpenAI truly believed that copying our news was fair and legal, it would not have concealed the fact that it had already done so.'
Plaintiffs' allegations go beyond hidden tools and databases. They also claim that OpenAI deleted billions of ChatGPT outputs after the lawsuit was filed, directly violating a court preservation order. Additionally, when submitting the 20 million chat record sample, OpenAI replaced some of the requested logs with other logs, further compromising the integrity of the evidence.
These actions have prevented plaintiffs from effectively proving that their copyrighted content was heavily copied by ChatGPT. Plaintiffs therefore ask the court to impose a series of sanctions: bar OpenAI from using the 20 million record sample as evidence; presume that ChatGPT logs would show extensive regurgitation and generation based on plaintiffs' content; bar OpenAI from arguing that the logs it provided do not show substantial regurgitation; and require OpenAI to cover plaintiffs' legal costs incurred in pursuing the evidence.
OpenAI spokesperson Drew Pusateri denied all allegations, saying plaintiffs are trying to invade the privacy of unrelated users after losing ground in the case. 'As the Times' case has weakened, they have been forced to drop some claims and now continue to try to invade the privacy of people unrelated to this case, including making these blatantly false allegations,' Pusateri said. OpenAI will continue to defend user privacy and the principle of fair use.
The lawsuit began in 2024, with the core dispute being whether OpenAI's use of copyrighted content from the Times and other news organizations to train GPT models constitutes fair use. Previously, OpenAI successfully persuaded the court to dismiss some claims, but plaintiffs have persisted. This sanctions motion is the latest escalation in the case and could have far-reaching implications for evidence rules and the future of AI copyright law.
This article is based on reports from TechCrunch and Fast Company, both third-party tech media outlets, which are based on court filings and legal statements. Key facts (such as Monaco's testimony, 78 million records, Project Giraffe) come from plaintiffs' court motion and have not yet been adjudicated by the court. OpenAI's denial is also included.
OpenAI is accused of hiding key evidence in a copyright lawsuit, including an internal search tool and 78 million chat records. If the court rules in favor of plaintiffs, it could overturn OpenAI's fair use defense and have a major impact on transparency in AI training data.
Primary report
Primary source
Also covered by 1 sources