NYT Sues OpenAI; Both Discover Business Model Actually Needs One

The New York Times, that scrappy 173-year-old startup, has advanced its landmark copyright lawsuit against OpenAI and Microsoft to a critical judicial phase, with all three parties presenting official arguments to a judge on Friday in hopes of securing a favorable ruling before trial. What began as a 2023 filing has now matured into the kind of legal theater that venture capitalists assured everyone would never actually happen—because surely disruption exempts you from basic property law. The case represents perhaps the most direct collision yet between the AI industry's build-first-ask-permission-later ethos and the legal system's quaint insistence that you cannot simply vacuum up decades of someone else's work product and call it training data.

OpenAI's entire competitive advantage rests on having trained its models on the internet's accumulated knowledge—which is to say, on material that journalists, authors, photographers, and musicians created, edited, fact-checked, and published under the assumption they retained some ownership stake in their own labor. The Times, naturally, took exception to its reporting being converted into a probabilistic text generator without compensation or consent, which raises an uncomfortable question: if your business model requires that lawyers eventually have a very expensive conversation with a judge, is it really a business model, or is it just a lawsuit with a chatbot attached? OpenAI and Microsoft apparently wagered they could build trillion-dollar companies on this bet. The Times apparently disagreed and filed suit.

This is not the first time technology companies have discovered that "disruption" and "copyright infringement" are not, in fact, the same thing. The peer-to-peer file-sharing industry made roughly this argument in the early 2000s. So did YouTube, before it made a peace treaty with the music industry. Google Books faced similar challenges before settling. The pattern is consistent: build a product that requires mass infringement, achieve scale, get sued, settle for a number that makes the lawyers rich and the business model technically "sustainable." What distinguishes OpenAI's situation is scale and stakes—we are not talking about a few million dollars in damages, but rather the foundational legitimacy of generative AI itself.

The courtroom arguments now being presented are, ostensibly, about arcane questions of fair use doctrine and the legal definition of transformative work. The Times argues that ChatGPT regurgitates its journalism verbatim in some cases—which, if true, would be difficult to defend as transformative. OpenAI and Microsoft will presumably argue that training is different from infringement, that large language models are inherently transformative, and that preventing them from using published text would chill innovation. Translation: the tech industry will argue that innovation means you get to use everyone's property until a court tells you to stop, at which point you negotiate a fee that was always going to be cheaper than licensing.

The critical vulnerability here is timing. If this case reaches trial, discovery could expose exactly how much Times content was in the training data, in what form, and whether it was deliberately included. If the judge rules for the Times, OpenAI faces either massive licensing bills or retraining costs—both existential threats to the current model. If the judge rules for OpenAI, it becomes legal precedent that your published work is available for commercial AI use without your consent, which would accelerate every competitor's race to vacuum up as much human-created content as possible before regulation arrives.

What matters most is not the legal outcome but what it reveals about the industry's actual business model: a bet that you can build something first, face lawsuits second, and settle third, with venture capital and access to cheap debt absorbing the frictional costs of doing business illegally. The Times v. OpenAI lawsuit is not a problem with AI—it is a problem with a company that thought the rules of copyright were a regulatory suggestion rather than, well, the actual rules.

When your business model requires you to hope a judge is more permissive than the law suggests, you do not have a business model. You have a liability.