NYT takes aim at Microsoft supercomputer
In 2023, the NYT became the first major publisher to sue OpenAI. The prominent newspaper alleged that ChatGPT illegally trained on its articles, infringed copyright by quoting the articles verbatim, and harmed the market by positioning ChatGPT as a replacement for NYT subscriptions, as well as defamation by falsely making allegations about the NYT report. In addition, the NYT alleges that ChatGPT outlets, which aggregated Wirecutter reviews, robbed their authors of commissions from lost clicks on affiliate links.
In the original complaint, the NYT discussed Microsoft’s supercomputing systems as if they were providing public cloud computing services. The updated complaint alleges that the supercomputer was specifically designed to help OpenAI infringe and that it was built to train AI on copyrighted works without permission. And, as the NYT argued, its articles were weighted more heavily by this system because both firms hoped to produce models of the highest quality journalism possible so that the level of writing could be confidently imitated in the output.
The NYT claims that by creating this “extraordinarily sophisticated” machine, Microsoft has not only helped identify infringing works, but also provided a means of illegally seizing copyrighted works.
“Microsoft designed it specifically to use the entire Internet — designed to give Times Works disproportionate exposure — to produce the most accomplished LLM in history,” the NYT claims.
And now it is alleged that he is making an unfair profit.
The NYT claims that “Microsoft’s deployment of Times-trained LLMs across its product line helped boost its market capitalization by a trillion dollars last year alone.”
Model outputs show market losses, NYT claims
Findings shared during discovery for the NYT, including a a large proportion of users’ ChatGPT sessions— remains some of the strongest evidence that OpenAI and Microsoft have created tools to replace the NYT by producing verbatim excerpts from their copyrighted work.
In some cases, users told ChatGPT that they tried to bypass paywalls and were able to see significant parts of articles by asking to see the “next paragraph”. In other cases, “models just spit out a few paragraphs.” To prove the market harm caused by the substitution, they shared examples in their complaint of side-by-side comparisons, as well as screenshots of the allegedly infringing results:
