Toronto-based artificial intelligence firm Cohere Inc. is asking US courts to abandon copyright lawsuits filed by news media companies, including the Toronto star, with publishers claiming that they misunderstood their business model and “deliberately misunderstood the tool.”
Publishers including Condé Nast, The Atlantic, and Politico filed a lawsuit against Cohere in New York in February. Publishers claim that Coleales is engaging in “large-scale, systematic copyright infringement” by cutting news content for AI models. The lawsuit included examples of Cohere’s AI tools replicate the article.
Cohere said in a motion to dismiss the lawsuit on Thursday that it was a “different kind of AI company.” Other developers, such as Openai, are building consumer chatbots used by hundreds of millions of people to help with homework, fiction writing and other tasks. Cohere said the tool is being used by companies for searching for their own data and translating documents.
“It suggests that one of the plaintiffs’ long complaints does not suggest that Cohere’s actual customers are using (or using) to infringe Cohere’s solutions,” according to a court filing.
Cohere said publishers will use demo tools for developers designed to showcase AI models to create examples of suspicious copyright violations. The tool itself is not a product, the company said. “Plaintiffs are intentionally constructing scenarios that don’t arise from actual use cases to shape their arguments,” the company wrote in a blog post. “This is a fundamental failure in which most of the plaintiff’s claims are destined.”
AI models are typically built by cutting huge amounts of content from the Internet without consent or payment to the creator of that content. Authors, artists and media companies have filed numerous lawsuits against AI companies that claim that this practice violates copyright law. Some publishers have also signed agreements with AI companies to license model training content.
Globe and Mail, along with CBC and other Canadian outlets, filed a copyright lawsuit against ChatGpt-Maker Openai in Ontario last year, claiming that it was “unfairly enriched” by illegally using news content to build products. San Francisco-based Openai has shown in court filings that it intends to challenge the lawsuit on the basis of jurisdiction.
Cohere’s appeal for rejection focuses on the output of the AI model. The company said the terms and conditions prohibit customers from using the product in a manner that violates copyright laws, and the litigation publisher has not argued that Cohere is liable for the user’s actions.
Media companies also argue that Cohere’s tools can provide a substantial summary of news stories. “The summary is not an infringement, especially when relaying facts that the summary is not protected,” the Cohere court filed.
When building AI models, Cohere said the goal is not to copy specific articles or documents, but to learn linguistic patterns with a large amount of text. The company noted whether the amount of fair use is a matter of lawsuits across the US.
Last year, Cohere and other developers said in consultations on copyright and AI that upholding copyright law exemptions allowed the federal government to build commercial models without being forced or licensed to use content creators’ materials.
