Unsealed Court Filings Reveal Microsoft and OpenAI Copyright Dispute Over News Training Data

Unsealed court filings in The New York Times' copyright lawsuit against OpenAI and Microsoft show Microsoft employees discussed whether using news articles to train models could amount to the largest labor theft in human history and trigger a “doom loop” that would degrade model quality. According to ChainCatcher, a 2023 internal Microsoft memo warned that millions of people worldwide would soon view large models “devouring” their work as an unprecedented theft and described large AI models as products that could destroy their own supply chain. Microsoft said the memos were written by Brent Hecht, its director of applied science, and did not reflect the company’s views. The company said Hecht was tasked with providing “different and asymmetric perspectives.” Microsoft CEO Satya Nadella testified that any content behind a paywall should be licensed by the party seeking to use it, and said Microsoft would have exercised its rights to require OpenAI to retrain its models if it had known in advance that paid content was being used. The filings also show an OpenAI employee told President Greg Brockman about building a “hack” to bypass The New York Times’ paywall, and Brockman replied, “Nice.” OpenAI and Microsoft both argue the training at issue qualifies as fair use. The lawsuit was filed by The New York Times in late 2023, and 11 publishers have since joined the case.