NYT Publisher Warns AI Giants Are Stealing Journalism to Build Competing Products

A.G. Sulzberger, publisher of The New York Times, stood before the World Media Congress and delivered a warning that has become impossible to ignore: the artificial intelligence industry is committing theft on a scale without precedent, and the future of journalism—and with it, the health of democratic societies—hangs in the balance.

The theft is systematic and brazen. AI companies scrape news websites without permission or payment, extract the reporting, repackage it as their own product, and use it to build chatbots and search tools that compete directly with the news organizations that created the work. This happens not once during training, but continuously, every single day. The justifications offered by tech executives shift constantly: innovation requires it, they say. Facts cannot be owned. Licensing takes too long. Fair use permits it anyway. Some even invoke national security, warning that if AI companies are forced to pay, America will lose the technological race to China. None of these arguments withstand scrutiny.

The scale of the dispossession is staggering. The six largest AI companies have a combined valuation of eleven trillion dollars—more than three times France's entire GDP. Private investment in AI in the United States reached nearly three hundred fifty billion dollars in 2025 and is accelerating. Yet less than half of one percent of that investment flows toward compensating the people and organizations whose work powers these systems. The New York Times alone published nearly half a million original works last year—articles, photographs, videos, podcasts—at a cost exceeding two billion dollars. The organization maintains journalists in all fifty American states and one hundred fifty-five countries. Some have worked from war zones, including more than seventy journalists and support staff deployed to Ukraine in 2025 alone. Over one hundred seventy-five years, the Times has produced twenty million original works. This is the material that AI companies have decided they are entitled to take.

Why do they want it so badly? Because it works. AI executives have admitted it would be impossible to train today's leading language models without copyrighted material. One engineer wrote that model success is determined by the dataset, nothing more. Five of the ten most-used sources for training popular language models belong to news publishers. The New York Times was the single largest source of proprietary data in one major training dataset, followed by other quality news organizations like The Guardian and the Los Angeles Times. When Microsoft's head of AI monetization acknowledged that "the web open was built on an implicit exchange of value," he was describing a bargain that AI companies have simply decided to break. They take the content, keep the revenue, and leave news organizations to collapse.

The damage is already visible. Over the past two decades, the United States has lost seventy-five percent of its journalists. More than three thousand newspapers have closed. A new one shuts down every three days. Entire regions of the country no longer have a single reporter asking questions at city hall or covering schools. The most expensive and difficult journalism—investigations, war reporting, rigorous accountability—has contracted most dramatically. Now AI threatens to accelerate this collapse. Traffic to major news websites has fallen more than forty-five percent on average as AI search tools intensify. Competing AI models send referral traffic at ninety-six percent lower rates than Google search. As people increasingly turn to AI chatbots for news and information, they bypass news sites entirely, depriving publishers of both readers and the advertising revenue that sustains them.

Many news organizations have turned to subscriptions to compensate for lost advertising revenue, which has fallen eighty percent over two decades. But as people discover they can access stolen journalism for free through AI products, building subscriber relationships becomes exponentially harder. Some larger news organizations have signed licensing agreements with AI companies, and others have negotiated micropayments for individual uses of their work. But there is little reason to believe these arrangements will generate enough revenue to offset what is being lost. Smaller news organizations, whose work has also been taken and used by AI models, have received no compensation at all. A survey found that the vast majority of publishers do not expect significant revenue from AI platforms.

Sulzberger's warning extends beyond journalism. AI companies have ransacked the entire corpus of human creative work—books, films, music, scientific research. Globally, creative professions employ more than fifty million people and generate approximately twelve trillion dollars in annual economic value. The New York Times has filed lawsuits against OpenAI, Microsoft, and Perplexity for flagrant violations of copyright law. But litigation is slow and expensive. The Times's case has already stretched across two and a half years and cost more than twenty million dollars. Most news organizations lack the resources to fight in court, which is precisely why AI companies have chosen this path.

The stakes extend far beyond business models. When local news organizations disappear, research shows that people in those communities begin to trust each other less and hate each other more. They become more isolated, less tolerant. Civic engagement declines and public corruption increases. A world increasingly mediated by AI systems that cannot be held accountable for accuracy, that distort news in nearly half of their responses, that refuse to correct errors, and that present falsehoods with unwarranted confidence is a world less equipped to sustain democratic self-governance. Two-thirds of Americans are deeply concerned about AI spreading misinformation, yet a growing share turns to AI for news anyway, sometimes trusting it more than the news organizations it depends on for its responses.

Sulzberger's message to news leaders gathered from more than sixty countries was clear: defend your intellectual property rights with legal action and legislative pressure. Negotiate carefully with AI companies, understanding that they already possess your content and intend to use it regardless. Unite with other creative industries facing the same predation. Pressure legislators to strengthen copyright protections rather than weaken them, to require AI systems to identify themselves and respect website restrictions, to demand transparency about how journalistic work is used, and to hold AI companies legally responsible for defamatory content they generate. Most importantly, news organizations must become destinations in their own right, building direct relationships with audiences rather than depending on platforms controlled by the very companies stealing their work. They must focus relentlessly on original journalism—the reporting, investigation, and analysis that AI cannot produce and that the public has nowhere else to find. In a world drowning in bots and digital garbage, trustworthy news and information have become rarer and more necessary than ever.