Internal Documents Reveal OpenAI and Microsoft Warned of 'Doom Loop' from AI Training Practices
Court documents unsealed in the ongoing New York Times lawsuit against OpenAI and Microsoft have surfaced internal communications that appear to contradict public positions taken by both companies regarding AI development practices.
The documents, which include internal emails and memos, reveal that employees at both companies were aware of potential harms from their training data practices. Microsoft's Director of Applied Science, Brent Hecht, notably characterized the companies' approach to scraping data for model training as the "largest theft of labor in human history" and a "complete mockery of the idea of fair use."
Internal Microsoft documentation reportedly described the situation as creating a "doom loop" that would damage the web ecosystem. The concern appears to center on the cycle where AI systems trained on internet data could eventually consume content generated by other AI systems, degrading quality over time.
Microsoft has sought to distance itself from Hecht's specific comments, with a spokesperson stating that the remarks do not represent the company's official position. The company has maintained that its use of training data falls within legal boundaries.
The unsealed documents are likely to feature prominently as the case proceeds, potentially influencing how courts weigh questions about AI training practices and copyright law. Both companies face ongoing legal challenges from content creators and publishers who argue that AI systems were built using copyrighted material without permission or compensation.