AI companies definitely can’t escape it. Creators and publishers never stop accusing them of plundering their works and content to train their models at low cost. This time, it’s Cohere that is at the heart of a complaint filed by several American, Canadian and English groups. Forbes, The Atlantic, Business Insider, Politico, Condé Nast and many others have joined forces.
Article theft and fake content
The plaintiffs have identified more than 4,000 articles — a non-exhaustive list — that were allegedly harvested without authorization by Cohere. In addition to training AI models, this content is also used to generate results. “Not only does Cohere steal our works, it also blatantly manufactures fake articles and attributes them to us, misleading the public and tarnishing our brands,” the complaint continues.
The publishers are seeking damages of up to $150,000 per infringing work (that’s $600 million or more), or an amount based on Cohere’s profits. The startup was valued at $5 billion in its most recent funding round. Other demands: that Cohere destroy all infringing copies of copyrighted works, and that a filter be put in place to prevent the system from scraping or copying copyrighted content.
Founded in 2019 by former Google researchers and based in Toronto, Cohere offers generative AI solutions for enterprises, with a focus on accessibility and cloud integration. Its models are designed for tasks like text generation, data analysis, and automating customer interactions.
Cohere calls the lawsuit baseless and “strongly” defends its “responsible” training practices for its AI models. “We have always prioritized controls to reduce the risk of intellectual property infringement and respect the rights of copyright owners,” the company told ArsTechnica.
0 Comments