OpenAI, Microsoft Knew of Web ‘Doom Loop’

Unsealed court documents in The New York Times' lawsuit against OpenAI and Microsoft reportedly show that the companies' own internal materials warned they were beginning a 'doom loop' that could damage the web. The filings also characterize large-scale scraping of data to train AI models as an unprecedented theft of labor, according to the summary. The documents are notable because they suggest the companies understood the potential harm to publishers and the broader web ecosystem while continuing to build their systems. At the center of the dispute is how AI models are trained. Publishers argue that their articles, books, and other creative work were used without permission or fair compensation. OpenAI and Microsoft are expected to argue that some use is permitted, but the internal warnings make the legal fight harder. If executives and researchers privately recognized a damaging cycle, plaintiffs will ask why safeguards, licensing deals, or opt-out mechanisms were not adopted sooner. A 'doom loop' describes a feedback loop in which AI-generated summaries reduce traffic to original sources, leaving fewer resources for quality journalism, which then makes the web less useful for future training and for users. The case could influence far more than one company. Courts may clarify whether training on copyrighted material is fair use, what transparency AI developers owe, and how damages should be calculated. Regulators could use the findings to push for data provenance, consent standards, and compensation models. Publishers, meanwhile, may demand collective licensing or technical standards that let them control how their content is crawled. The unsealed documents do not resolve the case, but they shift the debate. Instead of asking only whether AI training is legally allowed, they raise a question about corporate knowledge and responsibility. If the companies saw a threat to the open web, the public and the courts will want to know what they did in response. The outcome may shape how future AI systems source knowledge and whether the web remains a sustainable source of human-generated information.