NYT slams Microsoft for building copyright-infringing supercomputer for OpenAI - Ars Technica

Supercomputer clashes with Cox ruling? NYT slams Microsoft for building copyright-infringing supercomputer for OpenAI NYT shifts OpenAI/Microsoft copyright claims after SCOTUS ruling against Sony. 56 Credit: Bloomberg / Contributor | Bloomberg Credit: Bloomberg / Contributor | Bloomberg Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more Minimize to nav In a heavily redacted court filing Thursday, The New York Times proposed to amend its copyright complaint against OpenAI and Microsoft to clarify a claim and allege that Microsoft actively encouraged OpenAI to steal NYT works by building a bespoke supercomputing system ranked among the most powerful in the world. NYT’s motion comes after the Supreme Court sided with Cox Communications in a case where Sony tried and failed to claim that Cox was contributing to music piracy as an Internet service provider, which set a new standard for contributory infringement. Moving forward, plaintiffs will have to prove that parties intentionally acted to induce illegal conduct. Recognizing that the legal precedent has changed, the NYT now wants to amend its complaint to align its contributory infringement claim against Microsoft with that new standard. “Today, we asked the court for permission to file an amended complaint that further strengthens our case, clarifying our claim of contributory infringement against Microsoft based on new law and new evidence uncovered during discovery,” Graham James, an NYT spokesperson, said in a statement provided to Ars. In addition to clarifying one claim, NYT also agreed to voluntarily dismiss two claims of contributory copyright infringement and trademark dilution against all defendants. A Microsoft spokesperson told Ars that the company views the amended complaint as “a last-ditch effort by the plaintiff to save its claim from unfavorable precedent set in other recent rulings.” But in its motion, the NYT argued that neither Microsoft nor OpenAI would be prejudiced by allowing the amended complaint. It’s proper to allow plaintiffs to revise arguments when legal standards change, the NYT argued, and the case schedule would not be set back because “The Times does not seek any additional discovery in support of its amended claims.” “As we have long alleged, Microsoft actively encouraged OpenAI to steal our copyrighted works,” James said. “Beyond amending that claim and streamlining the case to its most potent arguments, our core claims remain the same from the day we filed this lawsuit—that Microsoft and OpenAI stole millions of The Times’s copyrighted works to compete with our products and illegally enrich themselves.” 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 on its copyrights by outputting articles verbatim, and caused market harms by positioning ChatGPT as a substitute for a NYT subscription, as well as reputational harms by falsely attributing claims to NYT reporting. Additionally, ChatGPT outputs 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 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, its articles were more heavily weighted by this system, as both firms hoped to train models on the highest-quality journalism possible, so that level of writing could be confidently mimicked in outputs. By building this “unusually complex” machine, Microsoft not only helped select the works that were infringed but also provided a means to seize copyrighted works without permission, the NYT alleged. “Microsoft specifically designed it for the purpose of using essentially the whole Internet—curated to disproportionately feature Times Works—to train the most capable LLM in history,” the NYT alleged. And now it’s allegedly unfairly profiting. “Microsoft’s deployment of Times-trained LLMs throughout its product line helped boost its market capitalization by a trillion dollars in the past year alone,” the NYT alleged. For the NYT, outputs shared during discovery—including a huge chunk of users’ ChatGPT sessions—remain some of the strongest evidence that OpenAI and Microsoft built tools that allegedly replaced the NYT by producing near-verbatim excerpts of its copyrighted works. In some cases, users told ChatGPT they were trying to skirt paywalls and were able to see significant chunks of articles by requesting to see the “next paragraph.” In other cases, “models simply spit out several paragraphs” without such finagling. To prove market harms caused by substitution, they shared examples in their complaints of side-by-side comparisons, as well as screenshots of allegedly infringing outputs: GPT and NYT side-by-side comparison. via NYT amended complaint GPT and NYT side-by-side comparison. via NYT amended complaint GPT and NYT side-by-side comparison. via NYT amended complaint GPT and NYT side-by-side comparison. via NYT amended complaint GPT and NYT side-by-side comparison. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Example of using GPT to skirt paywall. via NYT amended complaint Similarly as problematic for the NYT are hallucinations where Microsoft and OpenAI models falsely cite the NYT for content that they never published. The complaint listed examples like Bing Chat citing fake quotes from Steve Forbes’ daughter Moira Forbes and ChatGPT fabricating an NYT article that was never published but ChatGPT claimed linked non-Hodgkin’s lymphoma to consuming orange juice. “Users who ask a search engine what The Times has written on a subject should be provided with neither an unauthorized copy nor an inaccurate forgery of a Times article, but a link to the article itself,” the NYT alleged. Microsoft and OpenAI are hoping that the court will agree that training AI on NYT articles is fair use. In a statement provided to Ars, OpenAI spokesperson Drew Pusateri reiterated the AI firm’s often-repeated claims that AI training on copyrighted works is indisputably fair use. But the NYT likely expects that its evidence of substitution is strong, and that might not bode well for the tech firms it’s suing. Notably, one of the earliest verdicts finding that AI training was fair use was explicitly granted due to the plaintiffs’ failure to prove market harms. Last June, a federal judge laid out what he thinks could be a winning argument against AI training on copyrighted works, suggesting that the fair use question is far from answered. In this case, OpenAI has argued that “ChatGPT is not a substitute for a Times subscription,” the NYT reported, partly because “they transformed the material for a different use.” But if NYT manages to convince the court that the ChatGPT use is not so different from the newspaper’s, the most extreme outcome could require OpenAI and Microsoft to wipe models and start over. The NYT has also asked for permanent injunctive relief to prevent future infringement, as well as extensive damages, insisting that “as a direct result of their conduct, Defendants have wrongfully profited from copyrighted works that they do not own.”