NYT takes aim at Microsoft supercomputer
In 2023, the NYT became the first major publisher to sue OpenAI. The major newspaper alleged that ChatGPT was illegally trained on its articles, infringed its copyright by outputting the articles verbatim, and harmed the market by positioning ChatGPT as an alternative to a NYT subscription, as well as causing reputational harm by falsely falsifying claims about NYT reporting. Additionally, ChatGPT output summarizing Wirecutter reviews robbed writers of commissions from lost clicks on affiliate links, the NYT alleged.
In the initial complaint, the NYT discussed Microsoft’s supercomputing systems as if they were providing general cloud computing services. The updated complaint attempts to specify that the supercomputer was designed to help infringe OpenAI and alleges that it was created for the express purpose of training AI on copyrighted works without permission. And as the NYT alleged, its articles were given more weight by the system, because both companies expected to train models on the highest quality journalism, so that the level of writing could be confidently mimicked in the output.
The NYT alleged that by building this “unusually complex” machine, Microsoft not only helped select infringed works, but also provided the means to seize copyrighted works without permission.
The NYT alleged, “Microsoft designed it specifically for the purpose of using essentially the entire Internet to train the most capable LLMs in history—the Times Works has been curated to present itself as incoherent.”
And now it is allegedly making unfair profits.
The NYT alleged, “Microsoft’s deployment of Times-trained LLMs across its product line helped boost its market capitalization by a trillion dollars last year alone.”
NYT alleges model output harms market
For the NYT, the output shared during the search – which includes a large portion of users’ ChatGPT sessions – is some of the strongest evidence yet that OpenAI and Microsoft created the tools that allegedly replaced the NYT by producing almost verbatim excerpts of its copyrighted works.
In some cases, users told ChatGPT that they were trying to bypass the paywall and were able to view significant portions of articles by requesting to see the “next paragraph”. In other cases, “models easily spit out several paragraphs” without such closure. To prove the harm to the market caused by the substitution, they shared examples of side-by-side comparisons in their complaint, as well as screenshots of the allegedly infringing outputs:
<a href