The era of artificial intelligence announced its arrival less than four years ago with the public launch of ChatGPT. Within months, OpenAI’s chatbot collected 100 million users, making it the fastest-growing consumer product in history. Today, it is one of many increasingly powerful A.I. offerings, alongside those of Anthropic, Google, Meta, Microsoft and X.
There is little doubt that generative artificial intelligence represents the next major tech revolution, one that brings with it a dizzying array of important questions. Will A.I. spur a surge of productivity? Eliminate entire categories of jobs? Will A.I. unlock amazing medical breakthroughs? Facilitate a biological attack? Can the actions of A.I. models and agents be fully understood? Can they be controlled?
I’m here today to talk about questions that are, admittedly, somewhat narrower. But they matter a great deal to me, and to you and to society.
How will A.I. change the news? How will those changes affect the information ecosystem that serves as the public square for engaged citizens worldwide? And what can the people in this room do to ensure the future of first-hand, fact-based reporting that’s essential to the health of our democracies?
The early signs offer reasons for us to be concerned.
The companies driving A.I., already among the richest and most powerful in human history, are consolidating their outsize control over our data and our attention. At the same time, they are failing to embrace a core responsibility that comes with this power – to ensure the public has access to trustworthy news and information.
[SH]Brazen theft
Their hijacking of the public square is made possible by the original sin that animates their A.I. products – a brazen theft of intellectual property that has occurred at an unprecedented scale. Tech giants strip-mine news websites without permission or compensation. They repackage these stolen goods as their own, siphoning off the audiences and revenue that otherwise would go to the news organisations that created this work. And this happens not just once during the training process, but countless times every single day.
As a result, I fear we are careening toward a future with fewer and fewer journalists to do the expensive, difficult work of original reporting – going to places, talking to people, digging up information, covering important issues and events, providing context and analysis, investigating the powerful. A future where a crucial wellspring of a healthy society and a stable democracy – the truth, understanding and accountability provided by original journalism – continues to dry up.
This potential damage extends far beyond news. A.I. companies have raided civilization’s entire corpus of original works, an act that also poses a danger to the future of books, movies, music, research and an array of other fields. In the United States, these industries represent not just the heart of American cultural and intellectual life, but a pillar of its economy and one of its most powerful exports. Globally, creative professions employ over 50 million people worldwide who produce roughly $12 trillion of economic value a year.
The people gathered today lead news organisations from more than 60 countries. That means you have already fought through the gauntlet of pressures that have battered journalism everywhere, from disappearing revenues to technological intermediation to mounting attacks on press freedom. But on A.I., we must do more. Our profession has been too quiet, too passive and too fragmented in the face of abuses by the companies leading the A.I. revolution.
We cannot allow A.I. cheerleaders to dominate the public conversation without interjecting to argue for the importance of ensuring a sustainable future for original journalism. We cannot watch as A.I. companies attempt to permanently dismantle the rights that give us control over the work we create. We cannot sit by as this work is used to build replacement products that undermine our ability to earn the audience and revenue necessary to continue reporting the news.
Some tech leaders will portray my comments today as anti-AI. As defending the old status quo. As yet another ossified institution lashing out at the innovators who are driving the forward march of progress. And to be fair to our colleagues in Silicon Valley, there is a tradition of legacy incumbents – say a 175-year-old newspaper – complaining about new technologies and the disrupters behind them.
So it’s worth stating this plainly: The news organisation I lead, The New York Times, has a long record of embracing technology to advance the mission of independent journalism. We have a history of respectful partnerships with tech companies to bring that journalism to new readers in new ways. Meeting disruptions with curiosity, openness and adaptability helped us navigate the collapse of our print business and come out stronger on the other side. Today my colleagues are using A.I. technology – responsibly, ethically, and with humans making the decisions – to improve how we report, edit, distribute and monetise our journalism. Holding a powerful new technology at arms length is a recipe for failure.
And I fully believe A.I. has the power to do a great deal of good in the world. I’m not calling A.I. – or the tech giants that control this technology – inherently bad or evil. I’m warning that A.I. companies are making choices that violate settled law, threaten the viability of creative work and appear likely to cause a great deal of unnecessary harm.
News organisations should want the good A.I. can bring. But tech companies should also want to support the healthy, sustainable flow of the information and ideas and creativity that powers A.I. – to ensure their actions don’t lead us to a tragedy of the civic commons.
[SH]AI – the four basic ingredients
A.I. models are made with four basic ingredients.
The first is talent – the people who design the algorithms. The second is what tech companies call “compute.” That’s the infrastructure behind A.I., like chips and data centres. The third is energy, the electricity required to fuel these power-hungry products. The fourth is what tech companies call “data.” The word itself seems almost designed to make creative and expressive work sound trivial, a ubiquitous commodity. But “data” is often used, among other things, as a synonym for books, movies, music and journalism – what might more accurately be called “copyrighted content.”
Talent, compute, energy and data are all essential to the success of A.I. and, therefore, to the success of the tech giants.
The first three are paid for because – of course they are. No tech C.E.O. would dare suggest forcing the most talented engineers to work for free. To the contrary, they regularly offer pay packages worth tens, even hundreds of millions of dollars. Nor would they consider stealing chips from a Nvidia factory or illegally tapping a power line. Investors consider the potential financial rewards of A.I. to be so great that they are embracing losses that run into the hundreds of billions of dollars to build data centres and power plants.
In contrast, A.I. companies take “data” without consent or compensation. Their explanations for the theft keep shifting. They say innovation requires it. They insist they’re just taking facts, which no one can own. They complain that deals take too long and cost too much. They claim the “fair use” doctrine allows them to take content for free anyway. Sometimes they even invoke national security – they warn that if A.I. companies are forced to pay, America will lose the technology race to China.
None of these arguments withstand scrutiny. A chatbot can only spit out “facts” because it illegally copied entire news articles, which enables it to borrow just as liberally from protected language and style as well. Building data centres and power plants is far more expensive and time consuming than hiring lawyers to draft licensing deals with news organisations. Fair use doesn’t allow this kind of harmful, substitutive copying, retention and regurgitation of a single work, let alone everything humanity has ever produced. In its competition with China, America weakens itself by abandoning the intellectual property protections that fuel innovation and power America’s creative enterprises.
The combined valuation of the six leading A.I. companies is $11 trillion – more than three times the G.D.P. of France. Private A.I. investment in the United States reached nearly $350 billion in 2025 and is accelerating in 2026. So the theft of intellectual property is certainly not occurring for want of money to pay for it. Though publisher licensing deals are not public, given the small size of deals that have been reported, it appears that less than half of 1 percent of that investment is going to compensate the people and companies creating the data that powers AI.
Though there are many sources of data, A.I. executives themselves have recognised that original, high-quality content is particularly valuable for the effectiveness and reliability of the technology. Five of the top 10 sites used to train some of the most popular large-language models belong to news publishers. OpenA.I. confessed that it would be “impossible to train today’s leading A.I. models without using copyrighted materials.” One of the company’s engineers wrote that the success of models is “not determined by architecture, hyperparameters, or optimizer choices. It’s determined by your data set, nothing else.” In other words, you are what you eat.
[SH]The NYT’s experience
Let’s zoom in on the experience of The New York Times to see how this works.
If you want comprehensive and accurate answers in your A.I. chatbot, it’s hard to think of a better source of data than a news organisation that has for 175 years employed experienced, well-paid professional journalists to unearth new information, chronicle unfolding events and assess developments in politics, business, culture, sports, science and global affairs. This original work is valuable to tech companies in large part because it’s been carefully written and edited, independently verified, held to the highest standards for fairness and accuracy, and brought to life in a distinctive, compelling way.
Last year alone, The New York Times published nearly half a million such works, from articles to photos to videos to podcasts, at a cost of over 2 billion dollars. We put journalists on the ground in all 50 American states and in 155 countries, and those journalists more than occasionally experience life-threatening danger. In Ukraine, for example, we had more than 70 journalists and support staff on the ground. All of that was just in 2025. Stretch those contributions across 175 years and 20 million original works and you’ll get a fuller picture of what our newsroom has contributed to the public’s understanding of the world.
The distinctive value of Times journalism – as with other sources of quality journalism – has been repeatedly reaffirmed by the fondness that A.I. companies have for it. Although most A.I. companies conceal their training sources, The Times was the single-largest source of proprietary data in a major dataset used to train many different models, followed by a number of other news organisations, like The Guardian and The Los Angeles Times. A.I. companies regard pulling information from quality news organisations as one of the surest signs that their products are working correctly. As a Microsoft vice president said, “Premium content meaningfully improves response quality.”
Yet the tech giants have argued consistently that they should not be expected to ask permission to use – let alone pay for – this kind of intellectual property. Their argument, as their actions show, has been that they’re entitled to it. Meta trained its model on an infamous database of illegally pirated books. Perplexity openly defied the longstanding norm that websites can’t be scraped surreptitiously in defiance of their explicit objections. OpenA.I. has lobbied the American government to give it legal immunity from its seizures of other people’s work. Even Anthropic, often held up for its commitment to ethical A.I. development, has been unwilling to pay for the high-quality journalism it uses in its products.
