A federal judge has ruled that The New York Times can move forward with its copyright lawsuit against OpenAI and Microsoft, allowing the core allegations over AI model training to proceed toward trial.
The ruling, delivered on March 26, 2025, found that the newspaper’s central claims—that OpenAI used its journalism without authorization to train ChatGPT—are legally plausible. Some secondary claims were dismissed.
“We appreciate the opportunity to present a jury with the facts about how OpenAI and Microsoft are profiting wildly from stealing the original content of newspapers across the country,”Attorney Steven Lieberman, who represents The New York Times, said in a statement to NPR.
The case could influence how courts interpret copyright protections in the age of generative AI, particularly around whether scraping and summarizing news content for model training qualifies as fair use.
Allegations of Content Uso improprio e perdita di entrate
Il Times ha inizialmente intentato la causa il 27 dicembre 2023, a seguito di negoziati falliti con OpenAI. According to the complaint, OpenAI and Microsoft copied millions of Times articles to train large language models without seeking permission or offering compensation. The complaint outlines specific instances where tools like ChatGPT and Microsoft Copilot generated outputs resembling original Times reporting.
One prominent example involved product reviews from Wirecutter, the Times’ affiliate review site. The lawsuit alleges that AI-generated summaries mimicked Wirecutter recommendations and omitted affiliate links, leading to a loss in reader traffic and revenue.
The newspaper argues that these AI outputs function as replacements, not complements, for its journalism. “This is about replacing the content, not transforming it,”said attorney Ian Crosby during a January 2025 hearing in which the companies attempted to dismiss the case.
Microsoft and OpenAI Defend AI Model Practices
Both OpenAI and Microsoft argued in court that their training methods are protected under the fair use doctrine. Their legal teams drew historical parallels to older technology like photocopiers and video recorders—tools once challenged but ultimately deemed legal under copyright law.
Microsoft stated in its filings: “Copyright law is no more an obstacle to the LLM than it was to the VCR (or the player piano, copy machine, personal computer, internet, or search engine).”
OpenAI’s L’avvocato, Joseph Gratz, ha spiegato al giudice Sidney Stein che Chatgpt non replica interi articoli. Instead, it breaks source material into smaller components called tokens—units of text that enable the model to identify patterns and generate new responses.
“Regurgitating entire articles is not what it is designed to do and not what it does,”Gratz said. He added that The Times’ examples were derived from atypical prompts designed to elicit such responses.
The Times dismissed that defense, arguing that the functionality of AI models still poses direct harm. Crosby emphasized that these tools can divert as much as 30–50 percent of online news traffic from publishers’ websites.
Legal Costs, Dismissals, and What’s Left
By late 2024, the Times had already invested $7.6 million into legal expenses related to the case. Quasi $ 5 milioni sono stati spesi in un solo trimestre. The lawsuit seeks billions in damages and demands the destruction of any AI models trained on its content without a license.
In his March 26 ruling, Judge Stein dismissed the Times’ secondary liability claims but allowed the main allegations of direct copyright infringement to continue. His decision followed a hearing held in January, during which both sides debated the nature of AI training and the applicability of fair use.
Legal experts believe the case could set a precedent, potentially forcing AI companies to change how they gather and use training data. If courts side with the Times, developers may need to obtain licenses before ingesting journalistic content.
Beyond OpenAI: A Wider Legal Front
The Times’ lawsuit is part of a broader wave of litigation from publishers and creators concerned about the unlicensed use of their work in AI systems. Nel maggio 2024, otto giornali di proprietà di Alden Global Capital hanno presentato una causa simile contro Openi e Microsoft. Well-known authors, including Sarah Silverman and Michael Chabon, have also accused OpenAI of using their books without authorization to train ChatGPT.
Other publishers have opted for licensing over litigation. In 2024, OpenAI struck content deals with The Atlantic, Vox Media, TIME, News Corp, and the Financial Times, offering payment in exchange for access to their archives. These agreements offer an alternative model for AI training but also reveal the industry’s division over how to deal with content use in the AI era.
Perplexity AI also in the Crosshairs
In October 2024, The New York Times also sent a cease and desist letter to Perplexity AI, a smaller search-focused AI startup backed by Jeff Bezos. The letter alleged that the company summarized Times content without permission and ignored standard robots.txt directives used to restrict indexing.
Perplexity has denied using Times articles for training. Invece, afferma che i suoi modelli”indice”pagine Web disponibili al pubblico per le risposte basate sulle citazioni. La compagnia ha affrontato precedenti lamentele da Wired, Forbes e Condé Nast su questioni simili. Its CEO, Aravind Srinivas, has signaled a willingness to work with publishers and stated that “factual information shouldn’t be monopolized.”
Perplexity has also attempted to quell criticism by launching an ad revenue-sharing program and licensing partnerships with publishers such as Fortune and the Texas Tribune.
Lingering Questions About AI Training and Copyright
Openai ha affermato che intende offrire agli editori un maggiore controllo su come i loro contenuti vengono utilizzati nei set di dati di addestramento. Nel maggio 2024, la società ha promesso uno strumento”Media Manager”che consentirebbe agli editori di gestire l’inclusione nella formazione dell’IA. But as of now, that feature had not yet been delivered, drawing criticism from industry groups and small media outlets alike.
If the Times prevails in court, OpenAI and Microsoft could be required to rebuild their AI systems without any content scraped from the Times—an expensive and technically challenging prospect. Federal copyright law allows for fines of up to $150,000 per instance of willful infringement, meaning damages could scale rapidly depending on how many violations are proven.
For now, the case remains active in the Southern District of New York. Judge Sidney Stein’s March ruling sets the stage for deeper discovery and possibly a jury trial—one that could help define the future balance between AI development and content ownership. Gli osservatori legali stanno guardando da vicino, sapendo che le conseguenze possono estendersi ben oltre l’aula.