The Web's Doom Loop, Admitted in Court

Unsealed filings in the NYT vs. OpenAI lawsuit reveal executives knew LLMs were cannibalizing the very web they were trained on.

By The energy-climate Desk

For years, AI companies argued their model training was transformative—fair use, not theft. Now, unredacted court filings from the New York Times vs. OpenAI copyright lawsuit tell a different story, one the companies fought to keep sealed. [2]

An internal Microsoft document, cited in the filing, describes a self-defeating cycle: AI products scrape human-created content, cannibalize clicks from the sites they scraped, and destroy the business models that keep fresh content flowing. The document said: "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: 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.'" [2]

Microsoft executives, including CEO Satya Nadella, testified under oath that after ripping content from the New York Times and other news sites, clicks to those news sites fully cratered, falling by more than 90 percent on Bing. [2] Microsoft and OpenAI workers separately warned that publishers couldn't survive AI scraping, concerns that surfaced in the same proceedings. [1]

[2]

What makes the filing striking is not the novelty of the critique but the source—the companies themselves. Internal candor and public posture have rarely been so far apart.

The unredacted filing was found by Jason Kint, the CEO of Digital Content Next, a trade organization that represents digital media companies. [2] OpenAI and Microsoft's lawyers continue to argue fair use in court, even as their own documents describe a supply chain their products are actively consuming.

The largest theft of labor in human history.Brent Hecht