OpenAI, Microsoft docs warned of 'doom loop' for web

The unsealed documents come from the New York Times' ongoing copyright case against OpenAI and Microsoft. They reveal that, internally, both companies anticipated the damage their AI products could cause to publishers. Brent Hecht, Microsoft's Director of Applied Science, called the data harvesting the 'largest theft of labor in human history' and said Microsoft's fair use defense made a 'complete mockery' of the concept. A Microsoft internal document warned that their AI content strategy had started a 'doom loop' that would hurt both their models and the entire web.

Microsoft has since tried to distance itself from these statements, with spokesperson Alex Haurek saying they reflect one employee's individual perspective. In a separate filing, Jordan Usdan, GM for Data Strategy and Ops at Microsoft AI, described Hecht's role as adversarial and said he does not speak for Microsoft on these theoretical views. The filings also include statements from Satya Nadella, Sam Altman, and others.

OpenAI's own employees acknowledged that ChatGPT could reproduce copyrighted material verbatim, and that GPT-4 'memorized a ton of data.' OpenAI Policy Director Jack Clark warned about creating systems that substitute for human labor. The documents suggest both companies knew they might irreparably harm the publishing industry but proceeded anyway, and the court will now weigh whether that knowledge matters legally.