New York Times publisher accuses Big Tech of stealing journalism to power AI
The most expensive part of journalism may be the part AI makes look free. A reporter can spend three weeks chasing one fact: calling sources, reading court filings, sitting through meetings, waiting for someone to finally answer the phone. Then an AI tool can serve that same fact in three seconds, clean and confident, without showing the shoe leather behind it. That tiny gap is where the fight begins.
New York Times publisher A.G. Sulzberger told the World News Media Congress in Marseille on June 1, 2026, that AI companies are taking journalism at an “unprecedented scale.” The Reuters Institute noted in the same speech that ChatGPT reached 100 million users within months of launch, a speed few media products in history have matched.
At the same time, Pew Research Center says U.S. newspaper newsroom employment fell 57% from 2008 to 2020, dropping from about 71,000 jobs to about 31,000. That means this is not just a Silicon Valley story, or a New York Times story, or a lawsuit story.
It is a story about who pays for the facts that everyone wants to read, quote, summarize, scrape, and search for. Because if reporting becomes invisible, the bill does not disappear. It just lands on the people still trying to do the work.
What Sulzberger Actually Said
Sulzberger did not tiptoe around Silicon Valley. Speaking in Marseille, in remarks republished by the Reuters Institute, he accused major AI companies, including OpenAI, Google, Microsoft, Anthropic, Meta, and X, of building tools on journalism without permission, payment, or fair credit.
His sharpest line, “Their hijacking of the public square is made possible by the original sin that animates their AI products, a brazen theft of intellectual property that has occurred at an unprecedented scale.”
He also said tech giants “strip-mine news websites without permission or compensation,” a phrase that explains why publishers see this as more than a traffic dispute. The fear is simple: if AI gives readers the answer, fewer people click the original story, and the newsroom that paid for the reporting gets cut out of its own work.
The Legal Battlefield: Who Owns Training Data?
The New York Times had already taken that argument to court. Reuters reported that the Times sued OpenAI and Microsoft in December 2023, accusing them of using millions of Times articles without permission to train chatbots and create products that compete with the paper.
The lawsuit says the companies tried to “free-ride” on the Times’ investment, while OpenAI and Microsoft argue that training AI on copyrighted material can fall under fair use. The legal map is still being drawn.
Reuters reported in 2025 that Anthropic won a key ruling saying the use of books to train AI could be fair use, though the same judge said alleged use of pirated books raised separate problems.
Other cases, including claims against BloombergGPT tied to the Books3 dataset, have moved forward past early hurdles. So courts are not handing either side one clean victory. They are sorting one thorn at a time.
Why News Publishers Say AI Is Different
Publishers say AI is not just another search tool. Search engines usually send people somewhere. Chatbots can answer the question right there, calm and polished, with no obvious reason for the reader to visit the source.
The News/Media Alliance, which represents more than 2,200 publishers, argued in its 2023 white paper that generative AI developers copied and used news, magazine, and digital media content at scale. The group said some training datasets overweight publisher content by factors ranging from more than 5 to almost 100 compared with the broader Common Crawl web collection.
Danielle Coffey, the Alliance’s president and CEO, says, “Generative AI systems should be held responsible and accountable, just like any other business.” That is the publisher’s case in one sentence. If journalism fuels the machine, publishers want rules, labels, licensing, and money.
Big Tech’s Defense: Fair Use, Innovation, and Deals
AI companies and digital-rights advocates reject the theft frame. Their argument is that model training is closer to reading, indexing, and learning patterns than republishing full articles.
The Electronic Frontier Foundation warned in 2025 that a sweeping permission requirement could favor the richest firms and hurt smaller developers, researchers, and public-interest tools.
EFF lawyers Tori Noble, Mitch Stoltz, and Corynne McSherry wrote, “To work effectively, today’s GenAI systems need to be trained on very large collections of human-created works, probably millions of them.”
Yet the market tells its own story. Reuters reported in April 2024 that the Financial Times signed a licensing deal with OpenAI, allowing ChatGPT to show attributed FT summaries, quotes, and links.
OpenAI has also announced publisher partnerships with Condé Nast, Axel Springer, The Atlantic, News Corp, Vox Media, and others. If training is clearly free, readers may wonder why so many checks are being written.
Three Expert Perspectives Shaping the Fight
The fight now has three loud frames, each speaking to a different fear. Sulzberger’s frame is civic: if AI drains money from reporting, the public square gets weaker, and the Reuters Institute speech says he worries about “fewer and fewer journalists” doing expensive original work.
EFF’s framing is legal and social: a too-broad copyright wall could lock up learning, indexing, research, accessibility tools, and smaller AI projects. Coffey’s frame is economic: the News/Media Alliance says publishers are investing in journalism while AI firms collect users, data, brand power, and ad value from that content.
The U.S. Copyright Office’s 2025 report on generative AI training said the stakes are high and that lawmakers around the world are already weighing rules governing the use of copyrighted works in AI training. That is why this battle feels so charged. It is not just Times versus OpenAI. It is a fight over the price of knowledge in the age of instant answers.
The Numbers Behind the Power Struggle
The numbers explain why publishers sound so alarmed. Pew says total U.S. newsroom employment fell from about 114,000 workers in 2008 to about 85,000 in 2020, while newspaper newsroom jobs fell 57% over that span.
Northwestern Medill reported in 2025 that more than 130 newspapers closed in the prior year and that 50 million Americans now have limited or no access to local news. On the AI side, Sulzberger cited ChatGPT’s sprint to 100 million users within months, and the Reuters Institute’s 2026 Digital News Report found weekly use of AI chatbots for news rose globally from 7% to 10% of audiences.
That is still far from replacing news sites, but it is enough to make publishers sweat. A few percentage points in audience behavior can move millions of clicks, and millions of missing clicks can decide who hires a city hall reporter and who lets the desk go dark.
Niche Angles That Are Easy to Miss
The first overlooked angle is size. The New York Times has money, lawyers, subscribers, and bargaining power. A small local paper covering school boards, police budgets, and storm damage may not.
Medill’s 2025 report found 213 U.S. counties had no local news source, up from 206 the year before, while another 1,524 counties had only one. If AI licensing becomes the new toll booth, big publishers may get paid while small outlets get scraped or ignored. The second overlooked angle is that courts are moving in pieces, not one grand ruling.
Reuters reported that Anthropic won a fair-use ruling for training on books, but piracy claims remained a legal problem. The Times case is different because it argues that AI outputs can compete with fresh journalism, not just learn from older books.
The third angle is quite licensing. Deals with the Financial Times, Condé Nast, and others suggest a middle path, but middle paths often work best for those large enough to negotiate.
A Short Reflective Close
The old internet taught newsrooms that readers could disappear into platforms. AI may teach them something sharper: credit can disappear, too.
Sulzberger’s warning may sound fierce, but the fear behind it is easy to understand. Journalism is slow work in a fast-answer world. Someone still has to knock on doors, read court filings, call the mayor’s office, sit through the meeting, and check the facts that sound too neat.
If AI becomes the front door to information, the next question is not just who gets the answer. It is who paid to find it.
Key Takeaways
- A.G. Sulzberger accused major AI companies of “brazen theft” and “strip-mining” news websites during a June 1, 2026, speech at the World News Media Congress in Marseille.
- Reuters reported that The New York Times sued OpenAI and Microsoft in December 2023, accusing them of using millions of Times articles without permission.
- The News/Media Alliance, which represents more than 2,200 publishers, says AI developers have copied publisher content at scale and should pay for it.
- AI companies and digital-rights advocates argue that model training can be fair use, especially when it does not reproduce original works as substitutes.
- Licensing deals with the Financial Times, Condé Nast, News Corp, and other publishers show that a paid-content market is already forming, even as the legal fight continues.
- The outcome could shape how readers get news, how publishers make money, and how much original reporting survives in a world of instant AI answers.
Disclaimer – This list is solely the author’s opinion based on research and publicly available information. It is not intended to be professional advice.
Like our content? Be sure to follow us
