AI Copyright: US Newspapers Challenge LLMs in 2026

The class-action lawsuit recently filed by a group of US newspapers against major AI developers is a real turning point in the fight over AI copyright and content usage. This whole legal mess, which is about AI companies feeding copyrighted journalism into their large language models (LLMs) without permission, is forcing a serious conversation about what intellectual property even means anymore. Is this lawsuit going to completely change how AI models get their data?

For Sarah Chen, the managing editor at a digital news outlet in Phoenix, Arizona, the story began not with a bang, but with a slow, creeping realization. Her team at the Desert Sentinel had spent years producing local news, deep-dive investigations, and analysis for the Valley of the Sun, and that hard work, which costs real time and money, was the foundation of their subscription and ad revenue. Then, in late 2024, she saw something that made her stomach drop. AI news aggregators and chatbots were spitting out summaries and answering questions with phrases, facts, and structures that were just too close to her team’s original reporting. She’d look up a recent city council vote her team broke, and an AI would give a perfect summary of the Sentinel’s work, sometimes with zero credit or, even worse, a dead-end source link. “It was like watching someone copy your homework and then claim they invented the subject,” Chen said at a recent industry conference in San Francisco. Her frustration which is felt by editors and publishers everywhere, finally boiled over into action. This was about the fundamental devaluation of original journalism. The LLMs, trained on pretty much the entire internet, had hoovered up years of their copyrighted work without permission, payment, or any value in return. This is a direct challenge to the economic model that keeps independent news outlets alive. The class-action suit, filed in the Southern District of New York, points the finger at several big AI development firms. The plaintiffs, including the Desert Sentinel and other papers, argue that the defendants’ AI models were trained by swallowing millions of their copyrighted articles, photos, and videos without bothering to get a license or pay for them. The complaint says this is direct copyright infringement under 17 U.S.C. § 106, the law giving copyright owners the exclusive right to copy, share, and make derivative works from their content. Their argument is simple: an AI model that learns from and then mimics copyrighted text is making derivative works without permission. Period. Legal experts are glued to this case. Professor Eleanor Vance, an IP law specialist at the University of California, Berkeley, pointed out how tricky it is to prove infringement with generative AI. “The defense will undoubtedly argue fair use,” Vance told Reuters. “They’ll claim that training an AI model is far-reaching, a new use of the content that doesn’t compete directly with the original work. However, the plaintiffs will counter that the AI’s output often directly substitutes for accessing the original article, thereby harming their market.” This case goes beyond one company copying another’s code, it’s about the very raw material of knowledge being repurposed on a massive scale. The specific claims dig into how these AI models actually process data. When an LLM “learns” from an article, it isn’t just memorizing it, it’s breaking it down into patterns, grammar, facts, and writing styles. The plaintiffs argue that this process, and the new content it generates, leans so heavily on the original copyrighted work that it’s still an infringement, even if it’s not a word-for-word copy. Think about it: if an AI can summarize a complicated financial report that the Desert Sentinel spent a month investigating and published exclusively, and a user gets that summary from the AI, why would they ever need to visit the Sentinel’s website? That’s a direct hit to their subscriptions and ad money. The lawsuit wants both money for past infringement and injunctive relief, a court order that would force AI companies to stop using copyrighted material without a proper license. That second part is huge. A win for the publishers could fundamentally change how AI developers get their training data, possibly starting a whole new era of licensing deals and payments to creators. This is a critical moment for the content usage world. For Sarah Chen and her colleagues, this fight isn’t just about getting paid. It’s about whether quality journalism can even survive. “We invest in reporters, photographers, editors,” Chen said. “We send them to city hall, to war zones, to environmental disaster sites. If our work can be freely consumed and repurposed by AI without any return, how do we fund that essential work? It’s not sustainable.” Her fear is the same one echoing across the media industry, which has been fighting declining revenues and the ‘everything for free’ internet for years. Unchecked AI could just pour gasoline on that fire. The defendants, as you’d expect, are pushing back hard. Their main argument is that training an AI is a far-reaching process that falls under fair use. They say the AI creates something totally new from the patterns it learns, rather than just copying the original. They also claim that the sheer scale of the data they need makes getting individual licenses impossible, a claim publishers just aren’t buying. “Impracticality does not equate to legality,” the plaintiffs’ lead attorney shot back outside the Manhattan courthouse. “The law protects creators, regardless of the scale of the alleged infringement.” So, what is AI? A new kind of search engine, or a new kind of publisher? These legal proceedings are going to be a long, complex slog, likely taking several years. The final call will depend on how judges, and maybe eventually the Supreme Court, decide to apply old copyright law to this new generative AI tech. There’s no real precedent for this kind of AI training data case, so every ruling could be a landmark one. The stakes are massive, not just for the newspapers and AI firms, but for the entire digital economy and anyone who creates for a living.

One possible way out of this mess could be a collective licensing group, kind of like what ASCAP or BMI do for music. This group could negotiate blanket licenses with AI developers, making sure creators get paid fairly when their work is used to train AI models. A system like this would create a clear, structured way to handle content usage that actually pays creators while letting developers innovate. Without some kind of system, we’re just going to choke the human creativity that these AI models are supposedly built to mimic. And this isn’t just about news. The results here will ripple out and affect how book publishers, artists, and musicians think about AI copyright. If the newspapers can draw a clear line in the sand, it could help protect other creative fields. But if the AI developers win, we could be looking at a future where it’s almost impossible for human creators to make a living against machines that can produce similar content instantly and for practically nothing. Sarah Chen is pragmatic, but she’s holding out hope. “This is about ensuring AI develops responsibly,” she said. “It’s about recognizing the value of human creation and making sure the people who do the creating get paid. Otherwise, what’s the incentive to do the hard work of making something original and good?” She has a point. Protecting creators is what fuels innovation. This lawsuit by US newspapers against AI developers is forcing a much-needed re-evaluation of AI copyright and content usage, and it will end up defining the future for both journalism and artificial intelligence. The decisions from these courtrooms are going to set precedents for how we balance innovation with creation for years to come.