In a significant legal development, a US district judge in California has dismissed most of the claims against OpenAI by authors who accused the company of training its ChatGPT language model on pirated copies of their books. The judge, Araceli Martínez-Olguín, ruled that the plaintiffs, including notable authors like Sarah Silverman and Michael Chabon, did not provide sufficient evidence for their claims, except for direct copyright infringement, which OpenAI expects to defend against later in the proceedings.
The judge rejected the notion that every ChatGPT output is an infringing derivative work, stating that the plaintiffs failed to demonstrate substantial similarity between the AI’s outputs and the copyrighted books. Claims related to the Digital Millennium Copyright Act were also dismissed due to a lack of evidence that OpenAI intentionally removed copyright management information from the training data.
While most claims were dismissed, the judge allowed the case to proceed under California’s unfair competition law, acknowledging the possibility that using copyrighted works to train ChatGPT without permission could be considered an unfair practice. The authors have been given until March 13 to amend their complaints.
The outcome of this case could have broader implications for copyright law in the age of AI, as the US Copyright Office is currently seeking public input before issuing guidance on the matter. OpenAI has stated the necessity of using copyrighted materials for training AI models, emphasizing the need for current materials to ensure the relevance of AI outputs. The Copyright Office’s forthcoming reports are expected to have a significant impact on future legal claims and regulations.
Read more at Ars Technica…