The long-running copyright fight between news publishers and artificial intelligence companies has gained another major front.
A group representing 160 regional, local, trade and specialist publications owned by 26 publishers has filed a new complaint against OpenAI and Microsoft, accusing the companies of using copyrighted journalism without permission or compensation while developing their AI products.
The complaint was filed on September 16 in the U.S. District Court for the Southern District of New York. Publications involved include the Tampa Bay Times, Austin Chronicle and Florida Trend.
The case comes at a sensitive time for both companies. OpenAI and Microsoft are already fighting copyright litigation brought by The New York Times and other publishers. At the same time, recently unsealed court documents have drawn attention to internal comments about whether AI tools could replace some of the ways people traditionally use news websites.
That has pushed the debate beyond the question of whether copyrighted articles were used to train AI models.
The bigger issue now is whether generative AI is creating something genuinely new from that material, or whether AI-generated answers can sometimes act as a substitute for the journalism on which the systems were trained.
That question could become central to how U.S. courts eventually apply copyright law to generative AI.
What the publishers are claiming
The latest complaint alleges that OpenAI and Microsoft copied large amounts of copyrighted journalism while building and operating AI products.
According to reporting on the filing, the publishers say hundreds of thousands of articles were used without authorization in connection with products including ChatGPT and Microsoft Copilot.
Their argument is straightforward: publishers spend money employing journalists, editing stories and producing original reporting, while AI companies have allegedly benefited from that work without adequately paying the organizations that created it.
At this stage, however, these remain allegations.
The court has not ruled that OpenAI or Microsoft infringed copyright in the newly filed case.
Still, the lawsuit adds another sizeable group of publishers to a legal battle that has been steadily expanding.
Earlier in September, The Seattle Times and Newsday also sued OpenAI and Microsoft. Those newspapers allege that their websites were scraped, including material behind paywalls, and that their journalism was incorporated into data connected with AI products such as ChatGPT, Copilot and Bing's AI features.
They also argue that AI systems can reproduce or closely paraphrase their reporting and provide users with enough information that they may no longer need to visit the original news website or buy a subscription.
Those claims are also being contested.
OpenAI has said its models are trained on publicly available information and that its approach is grounded in the U.S. doctrine of fair use. Microsoft, responding to the Seattle Times and Newsday case, said it was surprised by the lawsuit while also acknowledging the importance of local journalism and indicating that it was willing to discuss possible solutions.
The fight goes back much further
The current wave of lawsuits is part of a much bigger copyright dispute.
One of the most important cases began in December 2023, when The New York Times sued OpenAI and Microsoft.
That lawsuit has since become one of the biggest legal tests of how existing copyright rules apply to generative AI.
Publishers argue that copyrighted journalism was copied on a huge scale to help create commercial AI systems. Their concern is not limited to what happened during training. They also argue that those systems can sometimes produce answers that compete with the original articles.
OpenAI and Microsoft reject that interpretation.
OpenAI's position is that training AI systems on publicly available information can qualify as fair use and that the process is transformative. The company says its models learn patterns and relationships from enormous datasets rather than simply storing a collection of articles and reproducing them on demand.
Microsoft has also argued that copyright law should leave room for transformative technologies.
Both sides have recently sought summary judgment in the broader litigation, asking the court to resolve important legal questions without sending every issue to a full trial. Judge Sidney Stein is overseeing the litigation in the Southern District of New York.
Why the newly revealed comments are getting attention
A new complication has emerged from documents and testimony disclosed during the litigation.
Recently unsealed material includes comments from OpenAI and Microsoft executives and employees discussing the possibility that AI products could substitute for news websites or publisher content.
Bloomberg Law reported that Microsoft CEO Satya Nadella testified that conversations with chatbots have "substituted" for publisher websites.
The report also said OpenAI executive Nick Turley had internally described the company's products as "largely substitutive."
Publishers are highlighting those statements because the effect of a new use on the market for the original copyrighted work is one of the issues courts examine when considering fair use.
But those comments do not, by themselves, decide the case.
Microsoft has argued that individual statements cited from internal discussions do not undermine its broader legal position. OpenAI and Microsoft continue to maintain that generative AI provides capabilities that are fundamentally different from simply republishing newspaper articles.
So the court still has to look at the wider evidence and decide how much legal weight those comments deserve.
Copyright law doesn't make every act of copying illegal
This is where the AI copyright debate becomes complicated.
Under U.S. law, copying copyrighted material does not automatically amount to infringement. Courts can consider whether a particular use qualifies as fair use.
Among the issues they examine are why the copyrighted work was used, the nature of the original material, how much was used and whether the new use affects the market for the original work.
For AI companies, the idea of transformation is particularly important.
OpenAI says AI models analyse enormous amounts of information to learn patterns that allow them to generate new responses. It has also pointed to recent federal decisions involving Anthropic and Meta in which judges found that certain uses of copyrighted material for AI training were transformative and qualified as fair use under the circumstances of those particular cases.
News publishers say their situation raises a different concern.
Their argument is that if somebody can ask an AI chatbot about a current event and receive a detailed answer based on reporting produced by a newspaper, that person may have less reason to visit the publisher's website.
If that happens at scale, publishers argue, it could affect website traffic, subscriptions and the emerging market for licensing journalism to AI companies.
The New York Times has raised similar concerns about AI-generated summaries reducing traffic to original articles. OpenAI disputes the evidence behind claims of market harm.
This is also why a ruling in one AI copyright case may not automatically settle every other lawsuit. Courts can look at how material was obtained, what kind of content was involved, what the AI system produces and whether there is evidence of damage to an existing or potential market.
Not every publisher is fighting AI companies
There is another side to this story that makes the relationship between AI companies and the news industry much less straightforward.
While some publishers are going to court, others are working directly with AI companies.
OpenAI has entered partnerships and programmes involving news organizations and journalism groups. The company has pointed to work with organizations including the American Journalism Project, the Lenfest Institute and WAN-IFRA as examples of ways AI technology can work alongside journalism.
Elsewhere in the industry, publishers have negotiated licensing agreements and technology partnerships with AI companies.
The result is an increasingly divided landscape.
Some publishers see licensing as a way to establish a commercial relationship with AI platforms. Others want courts to establish clearer limits on how copyrighted journalism can be used without permission.
The Seattle Times situation shows just how complicated those relationships can become. The newspaper has previously taken part in initiatives supported by Microsoft and OpenAI, yet it is now suing both companies over alleged copyright infringement.
Partnership and litigation, in other words, are not necessarily happening in completely separate corners of the industry.
The US government has entered the debate
The dispute has also attracted attention from Washington.
The U.S. Justice Department recently entered the broader legal argument with a court filing supporting the position that AI training on copyrighted text can, under certain circumstances, qualify as fair use.
The government described large-scale AI training as potentially transformative and connected AI development with wider questions of U.S. technological competitiveness and national security.
That does not decide the lawsuit.
The court must still examine the evidence and determine how copyright law applies to the specific conduct being challenged.
But government involvement shows how much is now riding on these cases.
The outcome could affect not only OpenAI, Microsoft and newspaper publishers, but also how future AI models are trained, how publishers negotiate licensing agreements and how people find information online.
What happens from here?
For publishers, the economic concern is easy to understand.
Original journalism costs money. News organizations pay reporters, editors, photographers and other staff to gather and verify information. Publishers argue that if AI systems can benefit from that reporting without paying for it — and then provide answers that reduce visits to the original source — the business model supporting that journalism could be weakened.
AI companies see the issue differently.
Their argument is that requiring individual permission for every copyrighted work involved in training could dramatically change how large-scale AI systems are built and could restrict forms of technological development they consider transformative.
The courts have not yet produced one clear answer that settles that disagreement across the industry.
That is what makes this latest lawsuit significant. It does not introduce an entirely new copyright argument. Instead, it shows that the dispute is spreading beyond a handful of major national news organizations.
Local, regional and specialist publishers are now increasingly part of the fight.
And underneath all these lawsuits sits one question courts may ultimately have to confront more directly:
If an AI system learns from journalism and can later answer questions that readers once went to a newspaper to answer, at what point does transformation begin to look like substitution?
For publishers and AI companies alike, the answer could shape the next phase of the digital news business.
