‘Doom Loop’: OpenAI and Microsoft Admit LLMs Is Destroying the Web, Built on Theft
microsoft openai
| Source: Mastodon | Original article
OpenAI and Microsoft admit their large language models are harming the web and are built on extensive content theft.
Executives at Microsoft and OpenAI have quietly acknowledged in internal documents that the very foundations of their large‑language‑model (LLM) products are built on what a senior Microsoft official described as “an astonishing theft of unprecedented proportions” and “the largest theft of labor in human history.” The language appears in court filings unsealed in the New York Times lawsuit against the two firms, where a Microsoft memo warns that the company’s AI content strategy has created a “doom loop” that harms both model performance and the broader web.
The memo explains that the aggressive scraping of publicly available text for training data not only raises copyright concerns but also fuels a feedback cycle: AI‑generated summaries replace original articles, siphoning up to 60 percent of referral traffic from news sites. As the web’s content is repurposed, the pool of fresh, high‑quality material for future models shrinks, degrading their usefulness while simultaneously eroding the revenue streams of publishers.
The admission matters because it provides the plaintiff with concrete evidence that the companies were aware of the detrimental impact of their data‑harvesting practices yet proceeded anyway. It sharpens the legal focus on copyright infringement, potential antitrust implications, and the broader societal cost of “predatory” AI systems that undermine the information ecosystem.
The case is still pending, and the next weeks will likely see intensified scrutiny from regulators and lawmakers in both the United States and Europe. Watch for further court motions, possible settlement talks, and any policy proposals that could force OpenAI, Microsoft and other AI developers to adopt more transparent, compensated data‑licensing models.
Sources
Back to AIPULSEN