Latest

Artificial Intelligence

More than 400 US newspapers demand the dismantling of ChatGPT in court

More than 400 US newspapers sue OpenAI and Microsoft for systematic theft of content and demand the dismantling of ChatGPT

More than 400 US newspapers demand the dismantling of ChatGPT in court
More than 400 US newspapers demand the dismantling of ChatGPT in court

More than 400 newspapers in the United States are seeking a court order compelling OpenAI and Microsoft to "remove all copies of registered works from all ChatGPT models and other large language models, as well as from all training sets."

If a federal judge were to rule that ChatGPT's foundations were built with stolen materials, the entire edifice of today's generative AI could be forced to be demolished and rebuilt from scratch, this time respecting the rules of the game. What is being debated in the courts is not a brake on technological progress, but a fundamental question for the survival of culture and information: does technical progress have the right to finance itself through the plundering and economic destruction of those who create knowledge?

An industry that has grown in the shadow of plunder

The gravity of this matter, formally filed on June 24, 2026, before the Southern District Court of New York, has exposed the seams of an industry that has grown in the shadow of plunder. A united front of 36 publishing groups, representing nearly 400 local and regional newspapers in the United States, has said enough is enough. Mastheads with more than a century of history such as The New York Amsterdam News, the Arkansas Democrat-Gazette and the Long Island Herald have decided to stand up to the two most powerful giants of Silicon Valley: OpenAI and Microsoft.

But this litigation transcends US borders and the format of the major dailies. The lawsuit describes with forensic precision a data-extraction model that strangles, with equal force, the base of the global news pyramid: thousands of small digital dailies, specialized blogs built on WordPress, such as the leading portal Observatorio Blockchain, and legions of independent content creators who are now discovering how the fruit of years of work has been devoured to feed someone else's business. Without their permission, and without receiving a single cent in return.

The engineering of expropriation

One of the most alarming pillars of the legal text lies in demonstrating that this siphoning of information was not accidental, but a conscious and planned business decision. According to the plaintiffs, OpenAI implemented automated software tools called Dragnet and Newspaper for the purpose of systematically stripping articles of their Copyright Management Information (CMI).

Through this technical maneuver, the algorithms separated the body of the text from any trace that would reveal its original authorship, deliberately removing the names of journalists and investigative reporters, the titles of the works and explicit copyright notices, and the editorial marks and terms of legal use.

By removing the labels from the merchandise, the tech companies fed the texts into their servers while concealing their provenance. The technical evidence submitted by the experts in the lawsuit is devastating. The C4 dataset, derived from the massive crawling of Common Crawl, held more than 115 million tokens belonging to the affected media outlets, while firms such as Ogden Newspapers accumulated more than 71 million tokens in the training corpus.

Identical words from copyrighted texts

The direct consequence is the phenomenon of memorization, since GPT-based models not only learn grammatical structures but retain literal fragments of the articles. The court filing notes that, given the appropriate commands (prompts), ChatGPT is capable of reproducing hundreds of identical words from copyrighted texts, evidencing automated plagiarism on an industrial scale.

The socioeconomic backdrop of the lawsuit reveals an obscene asymmetry. In March 2026, OpenAI's valuation rose to 852 billion dollars, underpinned by monthly revenue of 2 billion dollars and a critical mass of 900 million weekly users. Microsoft, far from acting as a mere financial investor, appears as the necessary co-architect. The company has injected 13 billion dollars and designed the supercomputing infrastructure indispensable for processing the stolen content and integrating it into commercial products such as Copilot and Bing Chat.

On the other side of this economic divide sits local journalism, a civil structure that sustains accountability in local communities. The plaintiff media outlets employ thousands of workers to cover the daily news that no artificial intelligence goes out into the street to gather. City council meetings, local elections, school boards, corruption cases in town halls, neighborhood budgets, crime reports, and local employment and economy.

Sam Altman's own admission

Silicon Valley chose to externalize the costs of research and writing onto publishers and creators in order to keep, exclusively, the profit from the automated answer. OpenAI's own CEO, Sam Altman, conceded the point to the plaintiffs when he declared before the British House of Lords that "it would be impossible to train the main current AI models without using copyrighted material." In other words, they admit the absolute need for the raw material, but refuse to pay the bill.

The impact of this extractive model alters the waterline of the entire open web. For more than two decades, the internet has operated under an implicit pact in which independent creators, analysts and journalists published their knowledge on the web in exchange for the possibility of receiving direct traffic, visibility, subscriptions or advertising revenue that would make their activity sustainable.

The irruption of generative AI has unilaterally broken that ecosystem. By diverting traffic, the user consumes the information summarized in the AI interface without ever clicking through to the original website. By monopolizing digital advertising, the big tech companies are drying up the well they feed from.

The Court's responsibility

If writing an in-depth technical analysis on a WordPress site or exposing a municipal corruption scheme requires weeks or months of effort and financial resources, but an algorithm appropriates the piece five minutes after publication to resell it in a chat, the production of original content becomes an unviable exercise. The Southern District Court of New York not only holds in its hands the resolution of a multimillion-dollar intellectual property dispute; it bears the historical responsibility of deciding whether the future of information will belong to legitimate creation or to unpunished plunder.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading