'The Largest Work Theft in History': Unsealed Documents Expose Microsoft's Internal Doubts

In the lawsuit filed by the New York Times, previously redacted documents reveal that a Microsoft executive described the massive collection of texts as early as January 2023 as a theft on an unprecedented scale.

The documents released in the proceedings initiated by the New York Times in 2023 against OpenAI and Microsoft had so far been widely hidden. Their lifting reveals the reservations expressed internally, as early as January 2023, by Brent Hecht, director at Microsoft, who described the massive recovery of texts to train models as a flight of staggering magnitude, and went so far as to speak of the greatest theft of work in the history of humanity.

The documents quantify the extent of the collection: 91,692 copies of articles from the New York Times, the Daily News and the Center for Investigative Reporting present in the training games, more than two million documents from nytimes.com in Common Crawl, and 160,903 separate works recorded in an internal corpus called Project Mango. The complainants also describe the circumvention of pay barriers, the deliberate removal of copyright statements and the use of Common Crawl as Bing's index.

On the economic side, the file reports a 93% drop in the number of clicks directed to the New York Times via Copilot. Nick Turley, Head of ChatGPT at OpenAI, spoke of an existential threat to publishers, while Satya Nadella acknowledged in her testimony that paid content should be licensed and ChatGPT replaced in practice the consultation of source sources.