A group of news publishers, including the Toronto Star, are suing Toronto-based AI company Cohere Inc. for copyright infringement.DADO RUVIC/Reuters
A U.S. court denied a motion from Toronto-based artificial intelligence company Cohere Inc. to dismiss a copyright lawsuit filed against it by major news organizations, including the Toronto Star.
The lawsuit filed in New York accuses Cohere of unlawfully scraping news content to build its AI models and violating copyright because its tools can allegedly reproduce articles word-for-word in some cases and provide substantial summaries, even if those articles are behind paywalls.
Cohere, which is worth US$7-billion, filed a motion in May asking the court to dismiss some of the claims, arguing that the publishers “deliberately misused” its tools to “manufacture” the case.
Judge Colleen McMahon rejected Cohere’s motion in a decision issued on Thursday. The ruling does not deal with the substance of the allegations, but whether the news publishers have adequately pleaded a case.
Cohere declined to comment but said in a statement earlier this year that the lawsuit is “misguided and frivolous.”
The news publishers include Condé Nast, The Atlantic, Forbes Media and McClatchy Media Company, among others. The Toronto Star is the only Canadian plaintiff.
“As the complaint alleges, Cohere has engaged in massive, systemic infringement, and the affected publishers deserve their day in court. This decision is the first step towards justice for these publishers,” said Danielle Coffey, president and chief executive officer of the News/Media Alliance, in a statement Friday. The non-profit group helped organize the lawsuit on behalf of its members.
How AI is reshaping the art and science of cooking
The case is one of many filed by publishers, artists and other creators against AI companies, which build models using copyrighted material without consent or payment. AI companies argue that such practices are allowed under copyright laws through a provision known as fair dealing or fair use, but how that applies to building commercial AI models has proven controversial.
Some AI companies have also struck payment deals with news organizations and others to train on their material.
While the lawsuit against Cohere makes claims about the use of copyrighted material in building AI models, the company had only asked the court to dismiss allegations that its tools produce material that violates copyright.
The news publishers identified 75 examples of alleged infringement by Cohere’s models that include verbatim text and another 25 that include a mix of copying and close paraphrasing, according to the decision. Cohere argued that the examples in the lawsuit differ in style, tone, length and sentence structure.
The court disagreed, writing that in some instances the output is “nearly identical” to the original works. “Publishers allege that Cohere designed its system to do exactly that. These allegations are sufficient to create a factual issue for jury consideration,” Justice McMahon wrote.
Cohere also argued that its AI tools are designed for businesses to improve productivity and that no customer would use its systems to intentionally violate copyright, writing that the news companies are “constructing a scenario that would not occur” in the real world.
Justice McMahon wrote that some of Cohere’s own marketing has promoted its tools as a way to keep up to date with the latest news and that the publishers “plausibly pleaded that Cohere takes affirmative steps to foster infringement.”
Another one of the publishers’ claims is that Cohere’s models will “hallucinate” false information and attribute it to a news outlet. Cohere argued that the lawsuit did not include a single real-world example of a consumer confusing a false AI-generated output with a legitimate news article.
Justice McMahon wrote that the news publishers were not obligated to do so at this stage and have “adequately alleged a likelihood of consumer confusion.”
Major news organizations including The Globe and Mail and the CBC are suing OpenAI in Ontario over copyright infringement. The ChatGPT maker had argued the Ontario court does not have jurisdiction over it and that the case should be dismissed. The court rejected that claim last week.