Unredacted Filings Show Microsoft Executive Called AI Scraping 'Largest Theft of Labor in Human History'

The documents, reported by The New York Times and cited by both outlets, provide internal reactions from executives at the two companies as they built large language models. In a January 2023 memo, Hecht called scraping an “astonishing theft of unprecedented proportions.” Microsoft’s own internal research found that Copilot could reduce click-throughs to The New York Times by as much as 93 percent compared with its conventional search engine, Bing, TechRadar reported. Hecht introduced the theory of a “doom loop,” and Engadget reported he said large AI models “are a product that destroys its supply chain.”

Turley, an OpenAI executive, said AI products are “largely substitutive” to journalism, according to Engadget. An OpenAI software engineer told colleagues in 2023 that “no matter how prominently we show the links, users won’t click,” the filings show. When an employee wrote to OpenAI president Greg Brockman about a new paywall hack, Brockman responded “ah nice,” according to the documents. Microsoft’s Hecht also warned that generative AI posed a “real risk” of disrupting the employment of people who unwillingly provided training data.

TechRadar reported that Microsoft CEO Satya Nadella admitted that had he known OpenAI was scraping data found behind paywalls, he would have exercised Microsoft’s rights to require OpenAI to retrain its models, excluding the protected data. The case accuses OpenAI’s mid-training datasets of including more than 91,000 pieces of work from The New York Times, Daily News and the Center for Investigative Reporting, according to TechRadar. OpenAI allegedly removed copyright notices from material before it entered the large datasets so that the models would not re-generate those notices for users.

Microsoft disavowed its employees’ statements. Spokesperson Alex Haurek told The New York Times that “Microsoft’s position is set out in its court filings, which explain why these transformative uses are consistent with copyright law and why Copilot is not a substitute for publishers’ journalism,” Engadget reported.

The Trump administration recently sided with AI makers such as OpenAI. A US Government statement of interest said, “Constraining LLM development under a misunderstanding of fair use doctrine would thwart such creative and scientific progress while hindering American prosperity and economic mobility.” The New York Times lawsuit, originally filed alongside five other writers, argues that Microsoft and OpenAI broke copyright law by scraping millions of stories and using the text without approval or compensation to train LLMs. Several similar lawsuits have already favored AI companies, but judges have said those rulings were made because the law around AI use has yet to be settled. The New York Times case remains open.