‘Doom Loop’: OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft

Executives from Microsoft and OpenAI have admitted that large language models (LLMs) are predatory technologies that have been built through "an astonishing theft of unprecedented proportions" by stealing content from human creators. This internal Microsoft document, cited in a New York Times copyright lawsuit against OpenAI, reveals that LLMs have created a "doom loop" that is killing the entire web and destroying the businesses these companies stole from.

According to the filing, AI products have caused a 90% drop in clicks to news sites after scraping their content. OpenAI and Microsoft executives have testified that their models steal content without providing economic value to the creators, threatening the livelihoods of millions around the world. Despite claims that their actions constitute fair use and transformative works under copyright law, whistleblowers argue that these technologies are fundamentally built on stolen content and pose an existential threat to the human labor market.