Court filings show OpenAI and Microsoft knew their scraping would harm the web

Newly unsealed court documents in The New York Times’ copyright lawsuit against OpenAI and Microsoft show internal warnings, written by the companies’ own employees, that their approach to gathering training data was going to damage the web. The filing runs 92 pages and quotes executives, researchers and policy staff describing risks they say the companies pursued anyway.

One internal Microsoft document, cited in the filing, warned that “Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time.” The document called it unusual for a product to threaten the economic foundation of the suppliers it depends on, and said that is what happened with the “content supply chain” behind large language models, the AI systems that generate text.

Microsoft’s Director of Applied Science, Brent Hecht, is quoted repeatedly. He described the harvesting of data for ChatGPT and Copilot as “the largest theft of labor in human history” and said Microsoft’s legal defense “makes a complete mockery of the idea of fair use,” the copyright doctrine that allows limited use of protected work without permission. Microsoft spokesperson Alex Haurek told The Verge that Hecht’s comments “reflect one employee’s individual perspective, are not a legal analysis, and do not represent the company’s views.” In a separate filing, Microsoft AI executive Jordan Usdan described Hecht’s role as intentionally adversarial, saying he is employed to bring “asymmetrical, futuristic, and academic” perspectives rather than to speak for the company.

The filing also quotes Microsoft CEO Satya Nadella acknowledging that chatbots have largely replaced the need to visit the sites that produced the information. Haurek said Nadella’s testimony addressed “broad principles and changes underway in how people find and consume information” and should not be read as a legal conclusion.

Traffic and memorization

According to the filing, OpenAI’s own media and economic experts linked declines in referral traffic for publishers, including the Times, to AI-generated summaries such as Google’s AI Overviews, and estimated search referrals could be down by as much as 60 percent. Separately, OpenAI employees are quoted acknowledging that GPT-4 “memorized a ton of data and therefore will be insanely good at regurgitation,” even as the company said preventing memorization was important to limit copyright violations. The filing cites instances of ChatGPT reproducing long passages from the Times, Mercury News, The Denver Post, Lifehacker and Eurogamer.

The document does not resolve the underlying copyright case. It presents excerpts selected by the Times’ lawyers to support their claims, and Microsoft disputes the conclusions being drawn from Nadella’s and Hecht’s remarks. What happens next depends on how the court in the Times’ lawsuit weighs these internal statements against OpenAI and Microsoft’s fair use defense, a question that remains unresolved.