Microsoft execs admit AI news is unleashing a "doom loop" on journalism in unsealed legal filing

In an unsealed legal filing, Microsoft executives have admitted that their AI news strategy has created a "doom loop" that is negatively impacting both the performance of their models and the entire web. The "doom loop" refers to the situation where the AI content threatens the economic foundations of its suppliers, in this case, the New York Times.

This situation has been unfolding for nearly three years, with the New York Times suing OpenAI and Microsoft for allegedly violating the newspaper's copyright. The lawsuit alleges that large language models like ChatGPT trained extensively on millions of New York Times articles without obtaining proper payment or licensing.

The executives' comments in the unsealed request for summary judgment reveal the severity of the impact. They describe the AI content strategy as having "hurt the performance of our models and the entire web at the same time." This statement highlights the unusual circumstance where an end-product is simultaneously detrimental to its own suppliers.

Furthermore, the filing points to the drop in Bing click-through rates on New York Times articles, which have plummeted by 83-93% since the widespread adoption of AI news summaries. This decline in usage demonstrates the self-defeating logic of the situation, as the AI models are effectively undermining the very source material they rely on.

The Times' lawyers argue that the copying of copyrighted material by OpenAI and Microsoft does not fall under "fair use" since the companies are essentially substituting their own products for the New York Times articles, resulting in significant commercial gain for the tech giants. The ongoing case continues to explore the complex issues surrounding copyright infringement in the context of AI-generated content.