Picture: Shutterstock/Feng Yu

Fourteen news and magazine publishers celebrated the “first step towards justice” in November with a first-round legal ruling in their favour against Canadian AI start-up Cohere.

It was the latest in a series of legal battles going on around the world pitting news publishers against AI companies which have taken their content without permission and used it to train their answer engines and image creation technology.

A German music rights society’s victory against OpenAI was also said to have set a precedent that would help protect journalists and their work.

The New York Times was the first major publisher to launch a lawsuit against an AI company, OpenAI alongside its partner Microsoft, in December 2023.

Although that case has not yet provided any precedent on whether the training of AI models on content without payment is permitted by law, several others are starting to build a legal picture.

Press Gazette will keep this page updated as new legal judgments on AI copyright are published.

You can find our rolling round-up of which news publishers are suing AI companies, and which are signing deals, here.

AI legal rulings around the world

14 news and magazine publishers in the US versus Cohere

• Win for publishers

*November 2025: *A judge dismissed Canadian AI start-up Cohere’s motion to dismiss a case against it brought by 14 news and magazine publishers in the US.

Judge Colleen McMahon, of the Southern District of New York, said the publishers had “adequately alleged” that outputs from the AI provider were “quantitatively and qualitatively similar” to their content.

The case involves Advance Local Media, Conde Nast, The Atlantic, Forbes, The Guardian, Business Insider, LA Times, McClatchy Media Company, Newsday, Plain Dealer Publishing Company, Politico, The Republican Company, Toronto Star Newspapers and Vox Media.

Judge McMahon said Cohere was entitled to republish the “underlying facts” in the publisher content but that publishers provided enough examples showing Command’s outputs had copied and pasted their work that it creates a factual issue for a jury to consider.

She said the publishers provided enough examples to show that “at least in some instances, Command delivers an output that is nearly identical to publishers’ works”.

The ruling also found that publishers had sufficiently alleged that Cohere “had actual or constructive knowledge” that its products were infringing their copyright.

Read the full Press Gazette story here and the legal decision here.

Getty versus Stability AI

• Win for AI

*November 2025: *Getty Images has failed to set a precedent banning AI training on its images in a copyright case against Stability AI in the UK.

Getty withdrew a key part of its case against Stability AI during the trial as it admitted there was no evidence the training and development of AI text-to-image product Stable Diffusion took place in the UK.

In addition a claim of secondary infringement of copyright was dismissed because, the judge said, “an AI model such as Stable Diffusion which does not store or reproduce any Copyright Works (and has never done so) is not an ‘infringing copy’” under UK law.

High Court judge Mrs Justice Joanna Smith did find in Getty’s favour that the use of Stable Diffusion in the UK sometimes generated images bearing Getty’s trademarks. However she said she had been unable to determine that this trademark infringement was “widespread” or continued in newer versions of Stable Diffusion after December 2022.

She said this meant Getty had no basis for a claim for additional damages.

Getty is still pursuing a similar case against Stability AI in the US. Mrs Justice Smith said the image company may be able to maintain its case “in the jurisdiction where the model was in fact trained”.

Read the full Press Gazette story here and the judgment here.

Thomson Reuters versus Ross Intelligence

• Win for publishers

*February 2025: *A US judge ruled it was not fair use for a competitor of Thomson Reuters to use its work to train an AI tool.

A summary judgment from US Circuit Judge Stephanos Bibas found summaries (known as headnotes) from Thomson Reuters’ legal information provider Westlaw were copyrightable because they put work into “distilling, synthesising, or explaining” non-copyrightable judicial opinions.

Ross Intelligence had made its own AI-powered legal research search engine by buying thousands of compilations of legal questions, which had been built by lawyers using headnotes from Westlaw.

Judge Bibas found that Ross infringed 2,243 summaries, saying “actual copying” in these cases was “so obvious that no reasonable jury could find otherwise”.

He also said Ross’s use was commercial and not transformative of the original work as it did not have a “further purpose or different character”.

Ross’s defences on “innocent infringement, copyright misuse, merger, and scenes à faire” grounds all failed but some issues will still go to trial.

However Ross has since appealed the decision, continuing to arguing what it did was fair use.

Trade association the News/Media Alliance expressed support for Thomson Reuters, saying: “Enabling AI companies to divert users away from legitimate subscription-based or ad-supported publisher sites to competing services under the guise of ‘fair use’ will destroy a rapidly growing market through which media outlets license website content and earn revenue from the AI developers that depend upon that content.”

It urged the appeal court to find that “the unauthorised exploitation of a copyrighted work to compete with the copyright owner’s own market is incompatible with fair use”.

Other creative industry representatives have also sided with Thomson Reuters, concerned about the precedent for AI training.

Read the full Press Gazette story here and the summary judgment here.

Relevant non-news AI legal judgments

GEMA versus OpenAI

• Win for AI

*November 2025: *German musical rights collecting society GEMA has won a case against OpenAI, with the AI company ordered to pay damages.

A German court ruled that OpenAI breached copyright law by training ChatGPT on popular artists’ song lyrics.

The ruling found that what ChatGPT was doing was not covered by the text and data mining exemptions under copyright law because the training process involved “not only extracting information from training data but also reproducing works”.

GEMA said the ruling was the “first landmark AI ruling in Europe” that “set a precedent that protects and clarifies the rights of authors: even operators of AI tools such as ChatGPT must comply with copyright law”.

The German Journalists’ Association said it “believes the ruling sends a clear signal that also extends to journalistic texts… Journalists who wish to take action against AI developers now have a stronger legal position.”

Read a summary of the judgment here.

Authors versus Meta

• Win for AI

*June 2025: *A US judge denied partial summary judgment for a group of authors against Meta, saying the company’s use of pirated books to train its Llama AI models was fair use and transformative.

US District Judge for the Northern District of California Vince Chhabria also found: “Meta has defeated the plaintiffs’ half-hearted argument that its copying causes or threatens significant market harm.”

The 13 authors involved included Richard Kadrey, Sarah Silverman, Ta-Nehisi Coates and Junot Diaz.

However the judge said he did not believe the case set a wider precedent. He said: “In cases involving uses like Meta’s, it seems like the plaintiffs will often win, at least where those cases have better-developed records on the market effects of the defendant’s use.

“No matter how transformative LLM training may be, it’s hard to imagine that it can be fair use to use copyrighted books to develop a tool to make billions or trillions of dollars while enabling the creation of a potentially endless stream of competing works that could significantly harm the market for those books. And some cases might present even stronger arguments against fair use.”

Read the full ruling here.

Authors versus Anthropic

• Win for AI

*June 2025: *A federal judge ruled it was fair use and “transformative” for Anthropic to use books to train its AI model Claude.

US District Judge William Alsup said Claude had “not reproduced to the public a given work’s creative elements, nor even one author’s identifiable expressive style”.

He also found: “The purpose and character of using copyrighted works to train LLMs to generate new text was quintessentially transformative.”

The lawsuit was brought by authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson.

Anthropic did infringe their works by creating and retaining a “central library”, however, the judge said. This opened the way to a trial but a settlement was reached in September so the arguments will not be heard.

Email [pged@pressgazette.co.uk](mailto: pged@pressgazette.co.uk) to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog