Microsoft's own leadership has admitted in internal documents that the company's artificial intelligence systems are caught in a self-destructive cycle, stealing content from news organizations while providing nothing in return, ultimately harming both the media industry and the AI models themselves. These startling confessions emerged from court filings in an ongoing copyright lawsuit against OpenAI by the New York Times, with Microsoft implicated as OpenAI's largest shareholder.

What Did Microsoft Executives Actually Say About AI and News Content?

In a particularly revealing internal document, Brent Hecht, Microsoft's director of applied science, laid out the mechanics of what he called a "doom loop." According to the unsealed court documents, Hecht wrote that Microsoft's "AI content strategy has started a 'doom loop' that will hurt the performance of our models and the entire web at the same time." He went on to explain the fundamental problem: "It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its 'content supply chain'".

Hecht

An LLM, or large language model, is the type of artificial intelligence that powers chatbots like ChatGPT and Microsoft's own Copilot. These systems are trained on vast amounts of text data scraped from the internet, including news articles, blog posts, and other published content. The problem, according to Hecht's analysis, is that AI companies are extracting this content without compensating the creators, while simultaneously making it less necessary for users to visit the original news websites.

"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," stated Brent Hecht, Microsoft's director of applied science, in internal documents revealed during the lawsuit.

Brent Hecht, Director of Applied Science at Microsoft

CEO Satya Nadella echoed similar concerns in his own communications, though in less dramatic language. According to the New York Times' legal filings, Nadella acknowledged that using AI chatbots "has substituted... giving you the information right there on the website on the AI platform versus needing to go to the underlying source." In other words, instead of clicking through to a news website to read an article, users can now simply ask an AI chatbot and get the information directly, bypassing the original publisher entirely.

Nadella

How Does This 'Doom Loop' Actually Work?

The mechanics of the doom loop are straightforward but devastating for news organizations. Here's how the cycle operates:

• Content Extraction: AI companies train their models on copyrighted news articles and other published content without permission or compensation to the original creators.
• User Behavior Shift: As AI chatbots become more capable, users increasingly ask them questions instead of visiting news websites directly, reducing traffic and advertising revenue for publishers.
• Economic Decline: With less revenue, news organizations have fewer resources to produce quality journalism, leading to less original content available on the web.
• Model Degradation: As the quality and quantity of original content on the internet declines, future AI models trained on this diminished content become less capable and useful.

This creates a vicious cycle where AI companies benefit in the short term by accessing free training data, but ultimately undermine the very ecosystem that makes their products valuable. The news industry loses revenue and resources, the web becomes filled with lower-quality content, and AI models trained on this degraded content become less capable over time.

The court documents also reveal that OpenAI showed little concern about the ethical implications of obtaining training data. When one OpenAI employee informed cofounder Greg Brockman about a method to bypass the New York Times' paywall without detection, Brockman's response was brief and approving: "ah nice".

What Does This Mean for the Future of AI and News?

The admission from Microsoft's leadership is significant because it represents an internal acknowledgment of a problem that AI companies have largely downplayed in public statements. While executives have publicly discussed AI safety and responsible development, these court documents show that at least some within Microsoft recognized a fundamental flaw in their business model: it depends on destroying the economic viability of the sources that make the AI useful in the first place.

The New York Times lawsuit is one of several legal challenges facing AI companies over copyright and content usage. These court filings suggest that the legal arguments against AI companies may have substantial merit, with the companies' own internal communications potentially serving as evidence of their understanding that they are harming the news industry.

For news organizations, these revelations underscore the urgent need for either legal protections, licensing agreements, or technological measures to prevent their content from being used to train AI systems without compensation. For AI companies, the documents suggest that sustainable AI development may require fundamentally rethinking how they source and compensate for training data.