The legacy media is marching over a cliff, waving the banner of intellectual property while whistling past its own graveyard.
When the New York Times and a coalition of mainstream publishers filed a motion asking a federal court to penalize OpenAI for allegedly destroying evidence, the media industry cheered. They treated it as a tactical victory, a David-versus-Goliath moment where the noble guardians of truth finally pinned the tech monster to the mat.
They are fundamentally misreading the board.
This lawsuit isn’t a defense of journalism. It is an expensive, short-sighted suicide note. By attempting to force generative AI companies into a legacy licensing framework, publishers are setting a trap that will catch no one but themselves. They are fighting a 20th-century legal battle against a 21st-century technological reality, and the collateral damage will be the independent press.
The Fatal Flaw of the "Stealing Our Words" Premise
The core argument of the publishing cartel rests on a fundamental misunderstanding of how large language models (LLMs) operate. If you read the legal complaints, you would think OpenAI has a massive, digital warehouse filled with pirated New York Times PDFs, ready to copy-and-paste articles on demand.
It is a comforting lie for an industry that desperately needs a scapegoat for its declining subscription revenues. But it is technically illiterate.
LLMs do not store text. They do not copy articles. They look at data to calculate the mathematical probability of how words relate to one another. When an AI model trains on an article, it builds a statistical map of language.
Imagine a human art student spending ten years looking at paintings in the Louvre. They memorize the brushstroke techniques of the masters, the color theory of the Renaissance, and the composition of classical portraits. When that student later paints an original masterpiece, they are not "stealing" the paint from Leonardo da Vinci. They are applying synthesized knowledge.
Legally, this falls squarely under the doctrine of Fair Use. The purpose of copyright is to prevent people from counterfeiting your work and selling it as their own. It was never intended to prevent a machine—or a human—from learning how information is structured.
By demanding that LLMs pay for every single byte of data they analyze, legacy publishers are trying to redefine reading as an act of copyright infringement.
The Accidental Monopoly: Why Big Media's Victory Hurts Everyone Else
Let’s play out the publishers' dream scenario. Assume the courts agree completely with the New York Times. The judge levies massive fines against OpenAI, rules that training on public data requires explicit, paid licensing agreements, and forces tech companies to pay billions to use news content.
What happens next?
The legacy giants pop champagne. For about five minutes. Then, the trap snaps shut.
The tech companies with trillions of dollars on their balance sheets—Microsoft, Google, Meta, Apple—will simply write the check. They will sign exclusive, multi-year, nine-figure licensing deals with a handful of elite media conglomerates (The New York Times, Axel Springer, News Corp). They will lock up the rights to the premium web, build a wall around it, and move on.
But what happens to the independent journalist? What happens to the mid-sized regional newspaper, the tech blog, the niche trade publication, or the substack writer?
They get completely wiped off the map.
An open web relies on the fact that any creator can publish content, get indexed by search engines, and be discovered by readers. If the legal precedent shifts so that data scraping requires massive up-front licensing fees, nobody will build an open AI model ever again.
Only the tech titans will possess the capital required to build and train AI systems. They will only license from the biggest media players. The long tail of independent media will be locked out of the AI ecosystem entirely. They won't get paid a dime, because the tech companies won't bother scraping them. They will be starved of traffic, starved of visibility, and starved of relevance.
I have watched digital media companies blow through hundreds of millions of dollars over the last two decades trying to pivot to video, pivot to Facebook, and pivot to algorithmic feeds. Every single time they cede control of their distribution to a tech monopoly, they lose. This lawsuit is the ultimate expression of that failure—voluntarily begging to be turned into a corporate content feed for an LLM cartel.
Dismantling the "Regurgitation" Myth
The plaintiffs in these lawsuits love to showcase examples of "regurgitation"—instances where a user prompts an AI model with a specific headline and the model spits out several paragraphs of a copyrighted article verbatim.
They present this as the ultimate "gotcha" evidence. It isn't. It's a manufactured edge case.
Regurgitation is a technical glitch known as overfitting. It occurs when a model is trained too intensely on a specific dataset, causing it to memorize instead of generalize. It is an error that AI engineers actively work to eliminate because a model that regurgitates text is actually a poorly functioning model.
More importantly, look at how these examples are generated. They require highly specific, adversarial prompts designed explicitly to break the model's guardrails. It is the digital equivalent of taking a crowbar to an ATM and claiming the machine is inherently designed for bank robbery.
In normal, day-to-day operations, users are not asking ChatGPT to read them the front page of yesterday's newspaper. They are asking it to analyze spreadsheets, write code, draft emails, and brainstorm ideas. The value of AI lies in its utility as a cognitive tool, not as a bootleg mirror of a newspaper subscription.
Stop Trying to Fix Copyright (Adapt Your Business Model Instead)
The real tragedy of this legal crusade is that it is a massive, expensive distraction from the actual problem: the legacy media business model is fundamentally broken, and it has been since Craigslist took away the classified ads section thirty years ago.
Publishers are treating AI as an existential threat to their copyright, when they should be treating it as a mirror reflecting their own lack of value.
If a user can get everything they need from a 50-word AI summary instead of clicking on your article, your article didn't offer much value to begin with. The internet was flooded with low-effort, high-volume SEO bait designed to capture programmatic ad cents. AI models can synthesize that commoditized information instantly because it requires no original thought.
The content that survives the AI era will not look like the content being fought over in courtrooms today.
- Original Reporting: AI cannot pick up a phone, interview a corrupt politician, or put its boots on the ground in a conflict zone.
- Deep Access: AI cannot build personal relationships with industry insiders to break exclusive news.
- Distinct Voice and Perspective: AI cannot replicate the unique, human point of view that makes a reader trust an individual writer over a generic corporate brand.
Instead of spending millions on white-shoe law firms to litigate a battle they will ultimately lose on appeal, media executives should be pouring that capital into hiring investigative journalists, building native platforms that don't rely on big tech algorithms, and creating membership models that people actually want to pay for.
The Hypocrisy of the Media Machine
There is a glaring, unspoken irony at the center of this entire legal drama.
For decades, news organizations have scraped the public web. They monitor social media feeds, quote public forums, reference academic papers, and aggregate the work of smaller local reporters without paying them a dime. They call it "fair use," "background research," or "the public domain."
When a human reporter reads twenty articles on a topic, synthesizes the information, and writes a twenty-first article, it’s praised as good journalism. When a machine does the exact same thing at scale, the Times calls it digital piracy.
You cannot claim the open internet is a free resource when you are extracting value from it, but a private property zone the moment someone extracts value from you.
The New York Times is trying to have it both ways. They want the traffic from Google’s search bots, but they want to sue OpenAI’s training bots. They want the benefits of an open web without the competition that comes with it.
The legal system will eventually sort this out, and history suggests it will not side with the protectionists. The printing press faced furious opposition from scribes. Recorded music was supposed to destroy live performance. VCRs were called the "Boston Strangler" of the movie industry by Hollywood executives.
Every single time a transformative technology arrives, the incumbent powers try to sue it out of existence. And every single time, they fail.
Quit crying to the courts. Fire the lawyers. Double down on real, un-hypeable, human journalism. Or get out of the way.
