OpenAI denies infringement allegations in author copyright cases

Aug 28 (Reuters) - OpenAI has responded in California federal court to allegations that it misused the work of authors including Michael Chabon, Ta-Nehisi Coates and comedian Sarah Silverman to train its artificial-intelligence language model.

The Microsoft-backed (MSFT.O) AI company said in an answer to the complaints on Tuesday that it makes fair use of copyrighted content to teach models like the one underlying its popular chatbot ChatGPT to create original material.

"The models learn, as we all do, from what has come before," OpenAI said in its filing. "The fair use defense exists for precisely that reason: to encourage and allow the development of new ideas that build on earlier ones."

Attorneys for the authors and attorneys and spokespeople for OpenAI did not immediately respond to requests for comment on the filing on Wednesday.

Copyright owners including writers, news outlets and music publishers have filed several high-stakes lawsuits against tech companies over the alleged exploitation of their work without permission in order to train text-based generative AI systems. The group of authors that includes Silverman, Coates and Chabon filed separate lawsuits against Meta Platforms META.O and Microsoft-backed MSFT.O OpenAI over their systems last year.

Meta and OpenAI have both convinced judges to dismiss some of the claims, though courts have not yet addressed the core question of whether the use of material scraped from the internet to train AI infringes copyrights on a massive scale.

Tech companies have said that AI training is protected by the copyright doctrine of fair use and that the lawsuits threaten the burgeoning AI industry.

Fair use promotes freedom of expression by allowing the use of copyright-protected works without prior permission under certain circumstances. Courts often focus on whether a use is transformative to determine if it is fair.

OpenAI said in its filing on Tuesday that its AI training is "paradigmatic transformative fair use."

"The process of training an AI model does not involve any communication of protected expression to a human audience," OpenAI said. "Instead, [its] purpose is to create new material that never existed before, based on an understanding of language, reasoning, and the world."

The case is In re OpenAI ChatGPT Litigation, U.S. District Court for the Northern District of California, No. 3:23-cv-03223.