A group of major North American media companies filed a lawsuit today against Toronto-based generative artificial intelligence (AI) startup Cohere, alleging that it engaged in “massive, systematic copyright infringement and trademark infringement.”
The consortium of publishers suing Cohere includes The Atlantic, Condé Nast, Forbes, The Guardian, Insider, the Los Angeles Times, Politico, the Toronto Star, and Vox, among others.
In their complaint, filed in the Southern District of New York, the publishers accuse Cohere of unfairly using their content to train its AI models and displaying full or partial copies of articles. “Left unfettered, such misconduct threatens the continued availability of the valuable news, magazine, and media content that publishers produce,” the document states.
A Cohere spokesperson described the lawsuit as “misguided and frivolous.”
The publishers’ complaint alleges that Cohere has scraped copies of their articles from the internet without permission or compensation and used them to power its AI services. It also claims that Cohere “blatantly manufactures fake pieces” and attributes them to publishers, “misleading the public and tarnishing our brands.”
The group is seeking damages from Cohere, including up to $150,000 per work infringed, and a court order preventing the company from using copyrighted works to train or fine-tune its AI models. The consortium has also requested a jury trial.
“Cohere strongly stands by its practices for responsibly training its enterprise AI,” a Cohere spokesperson told BetaKit. “We have long prioritized controls that mitigate the risk of IP infringement and respect the rights of holders. We would have welcomed a conversation about their specific concerns–and the opportunity to explain our enterprise-focused approach–rather than learning about them in a filing. We believe this lawsuit is misguided and frivolous, and expect this matter to be resolved in our favor.”
Some of the publishers suing Cohere have licensed their content to other large-language model (LLM) makers. This group includes The Atlantic, Vox, and Condé Nast, which have inked deals with Cohere competitor OpenAI.
This marks the latest of a series of AI copyright suits filed against tech companies by a group that includes other news outlets as well as authors, visual artists, and musicians. Whether defendants’ copying amounts to “fair use” could be the defining legal question.
The Toronto Star and other large Canadian news organizations launched a similar suit to this Cohere action against OpenAI last year, alleging that OpenAI used their copyrighted works to train the AI behind its popular chatbot ChatGPT without permission or compensation. The New York Times has made comparable allegations in a lawsuit of its own against OpenAI.
In other recent AI copyright news, Anthropic struck an agreement in a suit brought by music publishers over song lyrics, Meta lost a fight in its battle with a group of authors, and Toronto’s Thomson Reuters won its case against now-defunct legaltech startup Ross Intelligence.
RELATED: Major Canadian news orgs sue OpenAI for copyright infringement
Founded in 2019 by former Google researchers, Cohere builds LLMs that power chatbots and other generative AI applications. Unlike some of its rivals, which include OpenAI, Anthropic, Mistral, and Google, Cohere caters exclusively to businesses.
Last year, Cohere closed $500 million USD in Series D financing at a $5.5-billion valuation, making it one of Canada’s most valuable tech startups. The company counts a group that includes AMD, Export Development Canada, Fujitsu, Inovia Capital, Nvidia, Oracle, and Salesforce Ventures among its backers.
Cohere offers customers that build applications on top of its AI models indemnification from intellectual property infringement claims.
As part of a recent public consultation on copyright in the age of generative AI, Cohere, Google, and OpenAI backer Microsoft have advocated to the Government of Canada for a legal exemption to the Copyright Act to allow them to use copyrighted materials to train their AI models without having to pay copyright holders or obtain their permission, arguing that text and data mining for AI training does not constitute copyright infringement.
Feature image courtesy Web Summit via Flickr.