OpenAI, Microsoft internally warned of 'doom loop' for web

Brent Hecht, a Microsoft director of applied science, called the practice of scraping publisher content to train large language models "the largest theft of labor in human history," according to unsealed filings in The New York Times' copyright case against Microsoft and OpenAI. He also wrote that leaning on fair use would make "a complete mockery of the idea of 'fair use.'" The Verge drew the passages from 92 pages of court records that became public on September 17.

Another Microsoft note said the plain part: "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 filing frames chatbots as replacing, not routing to, the sites that supplied their training data.

The numbers are what will get lawyers' attention. The Times cites publisher click-through rates falling 83 to 93 percent once AI-generated answers appear above the links, and consultants engaged by OpenAI estimated total search-referral decline of "as much as 60 percent."

Named executives are quoted throughout. Nick Turley, OpenAI's head of ChatGPT, described the product as "largely substitutive" for publisher sites. Jack Clark, the company's policy director, wrote that OpenAI was "creating systems that substitute for the labor of the people that define the 'culture' of society." Greg Brockman, according to the filing, answered a paywall-bypass discussion with "Ah, nice." Satya Nadella, testifying under oath, described chatbots as substitutes for publisher platforms.

Two of the analysts we follow flagged the filing.