Microsoft's Own Scientists Warned of AI's Devastating Impact on Web Publishers

Internal documents from Microsoft reveal that the company's own scientists have referred to the use of large language models as 'the largest theft of labor in human history.' The phrase, coined by Brent Hecht, a director of applied science at Microsoft, was intended for internal eyes only but has now surfaced in court exhibits related to a copyright case against OpenAI and Microsoft.

The documents show that Microsoft's scientists were aware of the mechanism they were monetizing and described it as a 'doom loop' that would harm both their models and the web at large. They noted that large language models are 'a product that destroys its supply chain,' referring to how these chatbots remove the need for users to click through to websites, thus reducing revenue for publishers.

Microsoft's own data measured this effect, showing that Copilot's answer engine drove click-through rates on The New York Times' site down by as much as 93 percent against ordinary search. OpenAI's training data included over 91,000 copies of works from the Times, Daily News, and Center for Investigative Reporting, among others.

OpenAI executives were also aware of the impact, with Nick Turley describing products that are 'largely substitutive' for journalism as an 'existential threat' to publishers. The internal documents raise questions about Microsoft's defense in the copyright case, which relies on a fair-use argument.