The ends can’t justify the means.  

This was my major takeaway in reading the Summary Judgement cross-motions filed Friday in rapid fire order by the defendant Big AI companies (OpenAI and Microsoft) and the plaintiff news publishers led by The New York Times.    

The defendants, and their allies, want to tie their entire defense in this landmark copyright case to the inevitability of the AI-driven world. And those allies now include none other than the Trump Administration and its Department of Justice, which put its thumb on the scale last week in support of OpenAI. Their ends are national security. China, China, China. And their means are a radical revision of fair use doctrine that leans hard on the idea that AI is “transformative.”   

They argue quite passionately that the entire corpus of the open web must be available to train AI in support of the profits of our glorious tech leaders who are defending these United States.  But this bizarre over-reach is not defensible under copyright law. That’s not the way the law works. This tactic is tech industry lobbying and positioning masquerading as national security policy to circumvent the law. Fair use is an important but limited doctrine that must consider “the effect of the use upon the potential market for or value of the copyrighted work.” That means that copyright protects the incentive for original creation. And what will we have without that? 

In the case of The New York Times Company v. Microsoft Corporation, OpenAI, Inc. et al., the news plaintiffs begin their Summary Judgment argument exactly where this case should end: substitution. Citing the 2023 Supreme Court 7-2 Warhol opinion, which said that Warhol’s “Orange Prince” could be considered transformative art except that the commercial purpose it was used for harmed the very market of the original photograph. They remind presiding Judge Sidney Stein that the decision established that substitution is copyright’s “bête noire” and narrowed the idea that “transformative” is some magical word.

The plaintiffs are not alleging a single illegal act that is existence of the AI itself. They seek liability for unauthorized copying at each phase: data acquisition, model training, output grounding, output copying, and what appears through redactions to involve the dealing of copies between OpenAI and Microsoft. Each use case must be justified under copyright law. The mere existence of ChatGPT and/or Copilot as what the Trump administration contends is a necessary ends cannot wash their hands and legalize every step required to build and commercialize the now famous LLMs. 

Microsoft’s own arguments expose the issue. In a flurry of rhetoric in a well-written brief, Microsoft describes its AI tool (now called Copilot) as letting users have a “conversation with the internet about what it knows, how it knows it, and where to go to learn more.” But the Internet does not “know” anything. Journalists know. Reporters know. Editors know. Publishers know because they have invested in reporting, vetting, verifying, legal review, editing, distribution and building their brand trust day in and day out. In the case of The Times, they have done so since before the U.S. Civil War.  

That is why I look forward to the underlying evidence supporting plaintiffs’ allegations around OpenAI’s WebText scraper. Plaintiffs say this scraper emphasized webpage quality by prioritizing work that is curated/filtered by humans. That allegation alone may defeat the core argument that the entirety of the web is a single undifferentiated monolith, available to be swallowed whole in support of the transformation into the glorious AI. OpenAI sought human curation. Quality. Music to the ears of any premium publisher.  

The web isn’t a vast undifferentiated thing. It is not even an all-you-can-eat buffet.  

Yes, the web contains an extraordinary amount of user-generated content, personal posts, public domain material, academic papers, and public records. However, it also contains pirated works on one end and professional journalism on the other end. Engineers and researchers building the LLMs made a conscious decision when they chose to prioritize human involvement and original creation by targeting the very content copyright exists to protect. It’s obvious why: This original creative work is the very definition of value. And that creative value is what copyright exists to protect.  

Ultimately, the law says the defendants will only be able to defend their violation of copyright law through a demonstration of fair use. As an affirmative defense, they bear the burden of proving no harm to the market of the plaintiff publishers in the individual use cases of the specific work(s). Microsoft fails to meet this burden. Common sense fails it, too.  

And OpenAI’s proof there is no market harm is even thinner, as they point to The Times’ own use of OpenAI models and a claim that Times traffic has increased, as if that proves there can’t be any harm from OpenAI’s theft of their content. It quite clearly does not. An innovative publisher like The Times may use AI tech while still being harmed by the unlicensed copying that helped build it. Aggregate traffic growth across the company is the antithesis of the use-specific analysis required to defend the exploitation of each individual work. 

That is why Mike Davis’s strong critique of the Department of Justice statement of interest is important. Davis, a conservative lawyer and activist closely aligned with MAGA and the Trump legal movement, called on DOJ to withdraw its statement supporting OpenAI. He writes on FOX News that “the government’s stated position would let Big Tech oligarchs scrape, copy and commercialize every creative work in America without permission or payment,” adding that the national security argument is “un-American and absurd.” His point lands because the American way is not to subsidize powerful technology companies by stripping creators of rights. It is to build markets that reward investment and enforce property rights. 

OpenAI and Microsoft can pay for what they use. They are trillion-dollar giants paying for billion-dollar, highly-controversial data centers, extraordinary amounts of power and compute, and well compensated engineers. Premium content should not be the one critical input of this complex system treated as free simply because it’s available in digital form. 

On that note, the plaintiffs land their systemic point as a true prisoner’s dilemma –  something Congress and the Trump administration should actually be tuned to hear. Every AI company benefits if reliable journalism continues to exist. Capitalism and democracy benefit, too. But each has an incentive to take that journalism for free while hoping others pay enough to sustain the ecosystem. If courts bless that model, publishers will be forced to subsidize their own replacement. Substitution. 

Make no mistake, this case is not about stopping AI. I am certain that The New York Times and other DCN members will continue to lean into the future.  But it is about whether AI will be built on fair markets and sustainable incentives for original work. Substitution is copyright’s bête noire because it erodes the economics of creation. Here, it is not a side issue. It is the very point. And on Friday, the rule of law was on The Times’ side.