A federal judge in California recently dismissed aspects of separate lawsuits filed by comedian Sarah Silverman and others who alleged that OpenAI used their literary works without authorisation to train its widely-used chatbot ChatGPT.
Litigants include authors Michael Chabon and Paul Tremblay.
Judge Araceli Martínez-Olguín stated that authors from the three different lawsuits haven't provided enough evidence to support their claims, except for direct copyright infringement, reported Ars Technica.
OpenAI had previously argued this point when it filed a motion to dismiss these cases last August. At that time, OpenAI asserted its confidence in defeating the direct infringement claim at a "later stage" of the legal process.
The authors claimed that by potentially repackaging their original works as ChatGPT outputs, OpenAI's flagship chatbot acted as a sophisticated scam, breached copyright laws and state regulations against unfair business practices and unjust enrichment.
The judge dismissed several copyright claims, including allegations of vicarious copyright infringement.
She sided with OpenAI, agreeing that the authors' claim that "every" ChatGPT output "is an infringing derivative work" was not enough to prove vicarious infringement.
Proving it requires evidence showing that ChatGPT outputs are "substantially similar" or even "similar at all" to the authors' books, she said.
Martínez-Olguín wrote, "Plaintiffs here have not alleged that the ChatGPT outputs contain direct copies of the copyrighted books," and added, "Because they fail to allege direct copying, they must show a substantial similarity between the outputs and the copyrighted materials."
Authors also couldn't convince Martínez-Olguín that OpenAI violated the Digital Millennium Copyright Act (DMCA) by supposedly removing copyright management information, or CMI, like author names and work titles, from training data.
According to the judge, this claim failed because authors didn't provide evidence that OpenAI intentionally removed CMI or designed the training process to exclude it.
Moreover, examples of ChatGPT referencing their names suggest that some CMI might still be in the training data.
Martínez-Olguín, however, noted that because the state law has a broad definition of what qualifies as "unfair," it's conceivable that OpenAI's use of the training data "may constitute an unfair practice."
She stated that the remaining claims of negligence and unjust enrichment did not succeed because authors only accused intentional actions and didn't clarify how OpenAI "received and unjustly retained a benefit" from training ChatGPT on their works.
Authors have been instructed to combine their complaints and have until March 13 to revise their arguments and pursue any of the dismissed claims.
OpenAI has asserted its intention to continue utilising copyrighted content in its training data, stating that it's essential for training AI models due to the broad scope of copyright covering various forms of human expression.
コメント