Microsoft executive labels AI data scraping as massive theft in unsealed court filings
Just the facts

Microsoft executive labels AI data scraping as massive theft in unsealed court filings

Summary

Newly released documents in The New York Times lawsuit reveal a Microsoft executive called AI training practices “theft” and detail how OpenAI and Microsoft scraped paywalled content, prompting concerns over the impact on publishers.

Unsealed filings in a copyright lawsuit filed by The New York Times against OpenAI and Microsoft show that a senior Microsoft executive privately described the companies’ AI training methods as "theft." The documents also contain internal remarks from OpenAI leaders warning that their models could pose an "existential threat" to publishers and journalists whose work was used to train the systems.

According to the filings, the firms bypassed paywalls to collect large volumes of news content, stripped copyright notices from the data, and built extensive training datasets that included millions of articles from the Times and other outlets. Internal Microsoft presentations note that the Copilot answer engine reduced click-through rates to the Times website by up to 93 percent compared with traditional Bing search, a decline the company termed a “doom loop.”

Microsoft CEO Satya Nadella testified that paywalled material should be licensed for training and said he would have required OpenAI to retrain its models if he had known they had scraped such content. OpenAI’s head of ChatGPT, Nick Turley, wrote that publishers face an "existential threat" from the chatbot, describing it as increasingly "substitutive."

The filings reveal that OpenAI’s mid-training datasets contain more than 91,000 copies of works from the Times, Daily News and the Center for Investigative Reporting, while a Common Crawl-derived set includes over two million documents from nytimes.com. A Microsoft internal memo referred to the practice as an "astonishing theft of unprecedented proportions" and the "largest theft of labor in human history."

The court documents also detail collaborative data exchanges between the companies, including Project Taxi and Project Mango, which supplied each other with training material. OpenAI researchers reportedly discussed methods to circumvent the Times’ paywall, with one comment describing the approach as a "nice" hack.

Both companies declined to comment on the newly released materials.

Keep reading — for free

Create a free account to follow the news. No card required.