The NYT targets Microsoft’s supercomputer
In 2023, the NYT became the first major publisher to sue OpenAI. The prominent newspaper alleged that ChatGPT was illegally trained on its articles, infringed its copyright by publishing articles verbatim, and caused market damage by positioning ChatGPT as a substitute for a NYT subscription, as well as reputational damage by falsely attributing claims to NYT reporting. Additionally, ChatGPT results summarizing Wirecutter reviews robbed writers of commissions for missed clicks on affiliate links, the NYT alleged.
In the initial complaint, the NYT spoke of Microsoft’s supercomputing systems as if they were providing generic cloud computing services. The updated complaint seeks to specify that the supercomputer was tailor-made to help OpenAI infringe and allege that it was built for the explicit purpose of training AI on copyrighted works without permission. And as the NYT alleged, their articles were weighted more heavily by this system, as both firms hoped to train models in the highest quality journalism possible, so that that level of writing could be imitated with confidence in the results.
By building this “unusually complex” machine, Microsoft not only helped select the works that were infringed but also provided a means to confiscate copyrighted works without permission, the NYT alleged.
“Microsoft specifically designed it for the purpose of using essentially the entire Internet, selected to disproportionately include Times Works, to train the most capable LLM in history,” the NYT alleged.
And now he is allegedly benefiting unfairly.
“Microsoft’s implementation of Times-trained LLMs across its product line helped increase its market capitalization by $1 trillion in the last year alone,” the NYT alleged.
Model results show damage to market, NYT alleged
For the NYT, the results shared during the discovery, including a a large portion of users’ ChatGPT sessions— remain some of the strongest evidence that OpenAI and Microsoft created tools that supposedly replaced the NYT by producing near-verbatim excerpts from their copyrighted works.
In some cases, users told ChatGPT they were trying to bypass paywalls and were able to see important snippets of articles requesting to see the “next paragraph.” In other cases, “the models just spit out several paragraphs” without such fixes. To demonstrate the market damage caused by the substitution, they shared examples in their complaints of side-by-side comparisons, as well as screenshots of allegedly infringing products:
