Internal Admissions in New York Times Lawsuit
Newly unsealed court filings in the long-standing New York Times vs. OpenAI lawsuit have brought to light internal communications from Microsoft and OpenAI executives, revealing private acknowledgments that their generative AI products are built on unauthorized content and pose an existential threat to the digital ecosystem. The documents, cited in a motion for summary judgment filed by The New York Times, contradict public arguments often made by these companies regarding the nature of their model training and market impact, as reported by 404 Media.
The filings highlight what internal staff described as a “doom loop”—a cycle where AI models ingest human-created content, only to then replace the websites that produced that content by providing answers directly to users, thereby cutting off traffic and destroying the economic viability of the original publishers.
The “Doom Loop” and Economic Impact
According to the court filings, Microsoft executives found that their AI integration, specifically within Bing, led to a cratering of traffic to news sites. Internal documents noted that clicks to news publishers dropped by more than 90 percent after content was integrated into AI search results. One Microsoft document explicitly stated, “Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time.”
Executives further admitted that the current model of AI development relies on what one Microsoft document described as an “astonishing theft of unprecedented proportions.” The filings include internal discussions where OpenAI co-founder Greg Brockman reacted positively to an internal “hack” designed to bypass The New York Times paywall. Furthermore, OpenAI’s policy director, Jack Clark, acknowledged that the company was “creating systems that substitute for the labor of the people that define the ‘culture’ of society.”
Legal and Regulatory Stakes
The unsealing of these documents serves as a significant turning point in the ongoing litigation. While Microsoft and OpenAI have consistently argued in court that their training processes constitute “fair use” and are transformative in nature, these internal admissions suggest that leadership was aware of the destructive potential of their tools on the labor market and the broader economic stability of content creators. The case remains a critical focal point for media organizations and regulators examining the future of generative AI and copyright protections.