started · updated
Microsoft and OpenAI internal documents reveal AI data scraping concerns
Court documents released during The New York Times' lawsuit against OpenAI and Microsoft have revealed internal discussions regarding the massive scraping of web content to train artificial intelligence models. Brent Hecht, Microsoft's Director of Applied Science, described the practice in a 2023 internal memo as 'the largest theft of labor in human history' and an unprecedented scale of theft.
The filings highlight concerns about a 'doom loop' or 'vicious cycle' where AI tools like Copilot reduce traffic to original news sources—in some cases by as much as 93%—thereby depleting the revenue and quality of content that AI models rely on for future training. OpenAI leadership also reportedly expressed concerns that chatbots pose an 'existential threat' to publishers by replacing the need for users to visit original websites.
Microsoft has responded to these revelations by stating that Hecht's comments represent his personal opinion rather than the company's official legal stance. The documents also detail how datasets used for training included millions of documents from The New York Times and other media outlets.
Entities
Brent Hecht · Microsoft · Nick Turley · OpenAI · Satya Nadella · The New York Times
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 4 SOURCES] The mass collection of content for AI training is the largest theft of labor in human history. notepad.lv · businessmag.al · qubit.hu · 24.hu
- [● 3 SOURCES] OpenAI leadership privately acknowledged that chatbots pose an existential threat to publishers. businessmag.al · qubit.hu · 24.hu
- [● 2 SOURCES] Microsoft stated that Brent Hecht's comments reflect his personal view and not the company's legal position. notepad.lv · 24.hu
- [○ 1 SOURCE] One dataset used by OpenAI included over two million documents from nytimes.com via Common Crawl. businessmag.al
- [○ 1 SOURCE] Microsoft Copilot has caused click drops of up to 93% for The New York Times compared to traditional Bing searches. businessmag.al
- [● 4 SOURCES] AI systems risk creating a 'doom loop' where they reduce traffic to publishers, leading to less quality content for future AI training. notepad.lv · businessmag.al · qubit.hu · 24.hu
- [● 2 SOURCES] Satya Nadella admitted under oath that chatbots often make it unnecessary for users to visit original news sources. qubit.hu · 24.hu