Unsealed Filings Show Microsoft Executive Called AI Training Data Scraping 'the Largest Theft of Labor in Human History'
Newly unredacted filings in the NYT v. OpenAI and Microsoft copyright suit quote a Microsoft director calling AI scraping practices theft and OpenAI's ChatGPT head calling them an existential threat to publishers.
Overview
Newly unredacted material in the copyright lawsuit The New York Times brought against OpenAI and Microsoft three years ago includes internal statements in which a top Microsoft executive privately described the companies’ AI training practices as “theft,” while OpenAI’s own leadership said its AI models posed an “existential threat” to the publishers and journalists whose work trained them, according to TechCrunch. The unsealed filing is the latest escalation in the three-year-old lawsuit, in which The New York Times initially alleged the companies violated copyright law by training generative AI models on its content, according to TechCrunch.
As previously reported, the underlying litigation has already produced aggressive discovery rulings, including an order for OpenAI to turn over more than 100 million de-identified ChatGPT conversation logs. The newly unsealed material adds internal company communications and documents to the public record for the first time.
What We Know
According to TechCrunch, in a January 2023 internal memo, Microsoft’s director of Applied Science, Brent Hecht, called the companies’ AI training practices “an astonishing theft of unprecedented proportions” and “the largest theft of labor in human history.” Engadget’s own review of the filings independently confirms that Hecht made the “largest theft of labor in human history” characterization.
A separate internal Microsoft presentation written by Hecht in January 2024 described declining traffic to publisher websites as a “doom loop” that would “hurt the performance of our models and the entire web at the same time,” according to TechCrunch. The presentation cited Microsoft’s own data showing that its Copilot “answer engine” caused click-through rates for The New York Times’ domain to drop as much as 93% compared to traditional Bing search, according to TechCrunch. A Microsoft document quoted in the filing states, “It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its ‘content supply chain.’”
OpenAI’s head of ChatGPT, Nick Turley, wrote in internal communication that publishers face an “existential threat” from products like the chatbot, which are “largely substitutive” and “will get more and more substitutive as they get better,” according to TechCrunch. Engadget’s reporting independently confirms that Turley described AI products as “largely substitutive” to journalism. OpenAI President Greg Brockman described the models as “excellent at news,” according to TechCrunch.
Microsoft CEO Satya Nadella testified in a deposition that “anything that is paywalled should be licensed by anyone who wants to use it…for grounding or training,” and said that if he “had been made aware that OpenAI had scraped and trained on information that was behind a paywall,” he would have “invoked [Microsoft’s right to] require OpenAI to retrain its models,” according to TechCrunch. Nadella also agreed under oath that conversing with chatbots “has substituted … giving you the information right there on the website on the AI platform versus needing to go to the underlying source,” according to TechCrunch.
The filings describe how the companies allegedly obtained training content, including by bypassing paywalls. When OpenAI researcher Nick Ryder told Greg Brockman about a “hack to get around nytimes paywall,” Brockman replied “ah nice,” according to TechCrunch, a reply Engadget’s reporting also describes.
The filings also detail the scale of alleged copying. OpenAI’s mid-training datasets alone are said to contain more than 91,692 copies of works published by The New York Times, the Daily News, and the Center for Investigative Reporting, while a Common Crawl-derived dataset included more than 2 million documents from nytimes.com alone, according to TechCrunch. The filing states that “OpenAI delivered the entire GPT-3 training dataset to Microsoft, which Microsoft used to evaluate how to implement OpenAI’s models within its own commercial products,” and that “Microsoft similarly provided training data to OpenAI through initiatives called Project Taxi and Project Mango,” according to TechCrunch. The companies allegedly assembled the Project Mango data into a training dataset containing copies of at least 160,903 unique works from the news publishers, according to TechCrunch.
Steven Lieberman, counsel for the New York Daily News, said in a statement shared with TechCrunch that “the evidence revealed here for the first time shows that OpenAI and Microsoft knew that what they were doing was wrong,” according to TechCrunch.
Microsoft distanced itself from its own employee’s statements. Spokesperson Alex Haurek said that “Microsoft’s position is set out in its court filings, which explain why these transformative uses are consistent with copyright law,” according to Engadget. TechCrunch reported that OpenAI and Microsoft did not return its own requests for comment.
What We Don’t Know
Much of the newly public material comes from The New York Times’ own brief rather than the underlying exhibits, which remain sealed, and the quotes are presented in the brief without their original full context, according to TechCrunch. The court has not yet ruled on the underlying fair-use question in this case, and judges in other AI copyright disputes have been largely favorable to AI companies’ fair-use arguments, according to TechCrunch.