NYT accuses OpenAI of hiding evidence in ChatGPT copyright case

Key Takeaways

• NYT alleges OpenAI secretly maintained a database of 78 million de-identified ChatGPT conversations to track copyright infringement
• OpenAI reportedly implemented 'Project Giraffe' to detect and record content regurgitation in outputs after the lawsuit was filed
• The plaintiffs are asking the court to sanction OpenAI and accept as fact that ChatGPT logs would show major content regurgitation

The New York Times and The Daily News have filed new allegations claiming OpenAI deliberately concealed evidence in their two-year copyright lawsuit over ChatGPT's training data. According to court filings, an OpenAI engineer's deposition revealed the company maintained internal tools to search its training corpus and track copyright infringement, directly contradicting its prior claims that such searches were technically impossible.

What did OpenAI allegedly hide?

The case centers on testimony from Vinnie Monaco, an OpenAI data privacy engineer, during an April court-ordered deposition. Monaco allegedly revealed that OpenAI had already conducted internal searches of its training corpus for copyrighted journalism works. This stands in stark contrast to OpenAI's repeated arguments throughout the case that it lacked the technical ability to perform such searches.

More damning: Monaco's testimony allegedly disclosed that OpenAI had compiled a database of approximately 78 million de-identified ChatGPT conversations before the NYT even filed its lawsuit. The company was using this database internally to assess how much copyrighted material it was reproducing.

Shortly after the lawsuit was filed, OpenAI allegedly implemented a system called 'Project Giraffe,' which included a 'Bloom' filter designed to detect and record instances of content regurgitation in ChatGPT outputs. The plaintiffs argue this proves OpenAI knew it had a problem and was actively monitoring it while telling the court it couldn't search its own data.

The discovery process has been contentious

The plaintiffs originally requested a sample of 120 million chat logs. OpenAI negotiated that down to 20 million. When OpenAI finally submitted the sample last December, the court found it so heavily redacted as to be 'unusable.' The plaintiffs also allege OpenAI deleted billions of ChatGPT outputs after the lawsuit was filed, violating a court preservation order, and substituted millions of logs in the requested sample.

Throughout this process, OpenAI argued that producing chat logs would be technically burdensome and raise user privacy concerns because logs would need to be retrieved, processed, and de-identified. The 78 million conversation database, if it exists as described, suggests that de-identification was already standard practice.

“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 are the plaintiffs asking for?

The NYT and Daily News are requesting that the judge sanction OpenAI for allegedly withholding evidence and obstructing discovery. Their specific asks are significant:

• Exclude OpenAI's 20 million chat log sample as evidence, claiming it is unreliable
• Accept as fact that ChatGPT logs would have shown major regurgitation of the plaintiffs' content
• Bar OpenAI from arguing that its provided chat logs don't demonstrate substantial copying
• Order OpenAI to pay the plaintiffs' legal fees for pursuing this evidence

If the court grants these requests, OpenAI would essentially be forced to concede that ChatGPT reproduces copyrighted journalism. That's a potentially catastrophic outcome for a company whose entire business model depends on the 'fair use' argument holding up.

OpenAI fires back

OpenAI spokesperson Drew Pusateri denied the allegations and accused the Times of trying to access private user conversations as its legal case weakens. '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,' Pusateri said. 'We'll continue defending our users' privacy and the long-established principles of fair use.'

The privacy argument is interesting. OpenAI is positioning itself as a protector of user data while the plaintiffs allege the company has already been analyzing de-identified user conversations for years. Both things can technically be true, but the optics are poor.

Why this matters beyond the courtroom

This lawsuit has always been about more than the New York Times. It's the most significant legal test of whether AI companies can train on copyrighted material without permission or payment. OpenAI's fair use defense, if successful, would validate the training practices of every major AI lab. If it fails, the entire industry faces potential liability.

The evidence concealment allegations, if proven, could shift judicial sympathy away from OpenAI regardless of the underlying copyright questions. Judges do not look kindly on parties who appear to obstruct discovery. And the specific claim that OpenAI was tracking its own infringement while telling the court it couldn't makes the company look like it knew it was doing something wrong.

Frequently Asked Questions

What is the NYT's copyright lawsuit against OpenAI about?

The New York Times and The Daily News sued OpenAI in December 2023, alleging the company trained ChatGPT on their copyrighted articles without permission and that the chatbot reproduces their journalism in user outputs.

What is Project Giraffe?

According to plaintiff allegations, Project Giraffe was an internal OpenAI system that included a Bloom filter to detect and record when ChatGPT outputs regurgitated content. It was allegedly implemented shortly after the lawsuit was filed.

What is OpenAI's defense in the copyright case?

OpenAI argues that training AI models on publicly available content constitutes fair use under copyright law. The company also raised user privacy concerns about producing ChatGPT conversation logs.

What could happen if the court sanctions OpenAI?

If the court grants the plaintiffs' requests, it could accept as fact that ChatGPT reproduces copyrighted content, which would severely damage OpenAI's fair use defense and potentially expose the company to significant damages.

How does this case affect other AI companies?

This is a test case for the entire AI industry. If OpenAI's fair use defense fails, other AI labs that trained on copyrighted material could face similar legal liability.

Source: TechCrunch / Rebecca Bellan