Will the company really get away with it after several lawsuits?
A US federal court has made an ambiguous decision in a high-profile copyright infringement case centered on OpenAI, the developer of the revolutionary ChatGPT. Most of the lawsuits filed by a group of well-known writers were rejected, but now the authors have a second chance to review the claims and try their hand again in the fight for justice.
The lawsuit titled "Paul Tremblay and Others v. OpenAI" began in 2023, when novelists Paul Tremblay, Christopher Golden, Richard Cadry, and actress-comedian Sarah Silverman accused OpenAI of illegally using their works to teach large language models.
As evidence, the plaintiffs point to the high degree of accuracy with which ChatGPT can retell the content of their creative materials. They also noted that the model deliberately ignores copyright-related information —the responses do not include ISBN numbers or authors names-which, in their opinion, is an attempt to disguise the violation.
The authors are particularly unhappy that OpenAI did not contact them to discuss nuances and compensation. At the same time, neural networks create texts that are surprisingly similar to the original works — the plaintiffs consider this a serious threat to their income.
Federal District Judge Araceli Martinez-Holguin, who handled the case in Northern California, received a request from OpenAI to dismiss the claims in August.
In a new ruling published on Monday, Martinez-Holguin told the authors unpleasant news: "The authors did not provide a clear explanation about the nature of the artificial intelligence responses, nor could they prove that any of the results had a significant or at least some similarity with their works. In this regard, the court decided to dismiss the claim for indirect copyright infringement."
According to the judge, there is no evidence that information critical to copyright protection was deliberately removed from the training material. The absence of this data in the neural network responses is not a sufficient reason to assume that developers are masking their illegal actions in this way.
Claims for illegal business practices, fraud, negligence, and unjust enrichment were also dismissed.
However, the judge allowed the claim of unfair commercial practices to be considered.
"If we assume that the defendants actually used the plaintiffs copyrighted works to train their models for commercial purposes, then such behavior, in the court's opinion, can be considered unfair," Martinez — Holguin pointed out.
The authors were given a new opportunity to clarify their original arguments and file a new complaint before March 13.
A US federal court has made an ambiguous decision in a high-profile copyright infringement case centered on OpenAI, the developer of the revolutionary ChatGPT. Most of the lawsuits filed by a group of well-known writers were rejected, but now the authors have a second chance to review the claims and try their hand again in the fight for justice.
The lawsuit titled "Paul Tremblay and Others v. OpenAI" began in 2023, when novelists Paul Tremblay, Christopher Golden, Richard Cadry, and actress-comedian Sarah Silverman accused OpenAI of illegally using their works to teach large language models.
As evidence, the plaintiffs point to the high degree of accuracy with which ChatGPT can retell the content of their creative materials. They also noted that the model deliberately ignores copyright-related information —the responses do not include ISBN numbers or authors names-which, in their opinion, is an attempt to disguise the violation.
The authors are particularly unhappy that OpenAI did not contact them to discuss nuances and compensation. At the same time, neural networks create texts that are surprisingly similar to the original works — the plaintiffs consider this a serious threat to their income.
Federal District Judge Araceli Martinez-Holguin, who handled the case in Northern California, received a request from OpenAI to dismiss the claims in August.
In a new ruling published on Monday, Martinez-Holguin told the authors unpleasant news: "The authors did not provide a clear explanation about the nature of the artificial intelligence responses, nor could they prove that any of the results had a significant or at least some similarity with their works. In this regard, the court decided to dismiss the claim for indirect copyright infringement."
According to the judge, there is no evidence that information critical to copyright protection was deliberately removed from the training material. The absence of this data in the neural network responses is not a sufficient reason to assume that developers are masking their illegal actions in this way.
Claims for illegal business practices, fraud, negligence, and unjust enrichment were also dismissed.
However, the judge allowed the claim of unfair commercial practices to be considered.
"If we assume that the defendants actually used the plaintiffs copyrighted works to train their models for commercial purposes, then such behavior, in the court's opinion, can be considered unfair," Martinez — Holguin pointed out.
The authors were given a new opportunity to clarify their original arguments and file a new complaint before March 13.
