The effect is not that people believe false things; it is that they stop believing true ones.
AI companies extracted news content without permission or compensation, redirecting audience and revenue from publishers while training models that replace original reporting work. The Times spent $2B+ on journalism in 2025 alone; its lawsuit against OpenAI and Microsoft has cost $20M+ over 2.5 years, a burden most smaller news organizations cannot afford.
- New York Times spent $2B+ on journalism in 2025; lawsuit against OpenAI and Microsoft has cost $20M+ over 2.5 years
- Major newspapers saw traffic drop 45% on average as AI adoption accelerated; advertising revenue fell 80% over two decades
- Six largest AI companies valued at $11 trillion combined; creative industries employ 50M+ people globally, generating $12 trillion annually
- ChatGPT credited news sources in only 1% of responses in test of 2,267 questions about Canadian reporting
New York Times publisher A.G. Sulzberger denounces generative AI as built on unprecedented intellectual property theft, comparing tech giants' data practices to digital piracy and calling for unified industry response to protect journalism.
A.G. Sulzberger, the publisher steering the New York Times through its digital transformation since 2018, stood before the world's newspaper editors in Marseille last week and called what happened to their industry by its plainest name: theft. Not metaphorical theft. Not the kind you can negotiate around. Theft on a scale without precedent, he said—the "original sin" that animates every major AI product now reshaping how people find information.
The generative AI systems that have captured the world's attention—ChatGPT, Claude, Gemini—were built by extracting news content without permission or payment, Sulzberger explained to the assembled leaders of 99 news organizations spanning more than 60 countries at the World Association of News Publishers conference. The tech giants then repackaged that stolen material as their own, siphoning away the readers and advertising revenue that would have flowed to the journalists and editors who created it. This happens not once during training, but continuously, every single day. He compared the practice to Napster, the digital piracy platform that became the defining legal battle of an earlier internet era—except this time, the companies doing it are among the richest and most powerful in human history.
The numbers tell part of the story. The New York Times alone spent more than $2 billion on journalism in 2025, publishing nearly half a million pieces of original work—articles, photographs, videos, podcasts—all carefully written, independently verified, held to exacting standards of accuracy and fairness. The organization deployed journalists to all 50 American states and 155 countries. In Ukraine alone, more than 70 Times journalists and support staff were working on the ground. Over 175 years, the Times has produced 20 million original works. Yet when researchers tested ChatGPT's ability to credit sources in Canadian news stories, the system attributed its sources in just 1 percent of responses. None of its competitors exceeded 16 percent. The theft, in other words, comes without even the courtesy of acknowledgment.
Sulzberger's lawsuit against OpenAI, Microsoft, and Perplexity has already consumed more than $20 million and stretched across two and a half years—a burden most news organizations simply cannot absorb. Smaller outlets whose work was also taken and used to train these models have received no compensation offers at all. Current licensing agreements and micropayment schemes, Sulzberger argued, are likely insufficient to offset the damage. The real wound is visible in the traffic data: major newspapers tracked by Comscore have seen traffic plummet by more than 45 percent on average as the AI race intensified over the past four years. Publishers globally expect those declines to continue. Less traffic means fewer advertising opportunities, and advertising revenue—already devastated, having fallen 80 percent over two decades—remains crucial to funding the expensive work of original reporting.
The broader threat extends far beyond news. AI companies have appropriated the entire corpus of civilization's original creative works—books, music, films, research—a category that employs more than 50 million people globally and generates roughly $12 trillion in annual economic value. The six largest AI companies are now valued at $11 trillion combined, more than three times France's entire GDP. They are not stealing because they lack resources to pay. They are stealing because they can.
What worries Sulzberger most is what happens next. As fewer journalists do the expensive, difficult work of original reporting—traveling to places, talking to people, investigating claims, covering important events, holding the powerful accountable—the information ecosystem will hollow out. A society and democracy require a crucial source: truth. Instead, what emerges is an internet saturated with low-quality content and bots, where it becomes nearly impossible to know where anything comes from or whether it is real. People stop believing false things and start believing nothing at all. The effect is not confusion. It is nihilism. Already, Pew Research data shows Americans are simultaneously at peak internet access, peak smartphone addiction, and peak social isolation. The tech companies shrug and say it is not their problem.
Sulzberger called on the assembled editors to mount a unified response. The tech giants repeat the Silicon Valley mantra that "information wants to be free," he said, but that phrase has a second half, rarely quoted, from philosopher Stewart Brand: "Information wants to be expensive because it is so valuable." The right information in the right place changes lives. Professional news organizations must position themselves as the trustworthy alternative to the chaos, he argued. Journalism is valuable. The question now is whether that value will be drained away by technology companies or returned to the organizations that do this essential work. On that answer depends the future of democratic institutions themselves.
Notable Quotes
The original sin that animates AI products is brazen theft of intellectual property on an unprecedented scale.— A.G. Sulzberger, New York Times publisher
Information wants to be free, but the second part of that phrase—rarely quoted—is that information wants to be expensive because it is so valuable.— A.G. Sulzberger, citing philosopher Stewart Brand
The Hearth Conversation Another angle on the story
When Sulzberger says this is the "original sin" of AI, what does he mean by that exactly?
He means the foundational act of theft—the training on unlicensed content—is what everything else flows from. Every product, every capability, every dollar of value is built on that initial taking. It's not a side effect or a regrettable accident. It's the core of how these systems work.
But couldn't you argue that AI companies are just doing what search engines did—indexing and summarizing content?
That's the distinction Sulzberger keeps returning to. Search engines direct you back to the original source. They send traffic and readers to the publisher. AI systems absorb the content, reproduce its insights without attribution, and keep the reader in their own interface. The user never needs to visit the news site. The value extraction is total.
He mentions that most news organizations can't afford to sue. Doesn't that make the legal strategy almost pointless?
It does, which is exactly why he's calling for collective action. One organization suing is expensive and slow. But if the entire industry moves together—if they coordinate licensing demands or refuse to allow scraping—they have actual leverage. Right now, each outlet is isolated and desperate.
What's the connection he's drawing between AI and democracy?
Original reporting requires resources: sending journalists to dangerous places, spending months investigating, maintaining editorial standards. If that work disappears because the revenue can't sustain it, then nobody is doing the hard work of holding power accountable. You end up with a public sphere full of noise and no reliable information. Democracy requires citizens to know what's actually happening.
Is he saying AI itself is bad, or just the way these companies are using it?
He's careful about that distinction. He's not arguing against the technology. He's arguing that the way it was built—by taking without permission or compensation—created a system that actively undermines the institutions that produce the reliable information the AI itself depends on. It's parasitic. It's eating the thing that feeds it.
The nihilism point—people not believing true things—that seems like the real fear underneath everything.
Yes. It's not just that misinformation spreads. It's that when everything is suspect, when you can't trust anything, people stop trying. They disconnect. And that's the endgame he's warning about. Not a world of lies, but a world where truth becomes meaningless.