OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft (www.404media.co)

In a revealing court filing related to the New York Times vs. OpenAI copyright lawsuit, executives from Microsoft and OpenAI openly admitted that large language models (LLMs) are fundamentally built on what they described as "an astonishing theft of unprecedented proportions." Internal documents highlighted a "doom loop" created by generative AI technologies that threatens to dismantle the entire web by cannibalizing traffic from the very creators whose content it relies upon. This candid acknowledgment calls attention to the ethical concerns surrounding LLM training, as it implicates these technologies in the erosion of traditional business models and wreaks havoc on the emotional and economic fabric of media creation. The statements from Microsoft and OpenAI executives underscore significant implications for the AI/ML community, as they challenge the prevailing narrative that AI can be built without considering the rights and contributions of human creators. Executives admitted that LLMs have led to dramatic declines in website traffic for news outlets, highlighting a dangerous cycle where AI systems not only rely on but actively devalue the work of journalists, artists, and creators. This filing marks a critical moment in the ongoing discourse about AI ethics, copyright, and the future of cultural production, as it forces the industry to confront the fundamental risks posed by technologies that, rather than enhancing creativity, threaten the existence of creative labor itself.