Federal Judge Araceli Martinez-Olguin dismissed significant portions of Sarah Silverman and several other authors’ lawsuit against OpenAI — the creators of ChatGPT. On February 12th, US District Judge Martinez-Olguin refused to allow certain claims that the authors had made in the complaint.
The federal judge’s decision echoes those made by another judge in California dealing with a similar case against Meta.
Martinez-Olguin dismissed one of the major claims that the authors had made: that all the answers ChatGPT generates are infringing work made of information obtained through copyright violation. Basically, everything ChatGPT creates violates the rights of the copyright holders of the works the company used to train its AI system.
The Details of the Case
Last year, Sarah Silverman, backed by authors like Michael Chabon and Ta-Nehisi Coates, sued Artificial Intelligence firm OpenAI. Last month, the presiding judge dismissed the claims of negligence, unjust enrichment, Digital Millennium Copyright Act violations, and vicarious copyright infringement. The authors were given until March 13 to amend their complaint and return to court. At the time of this writing, no public report has revealed whether or not they did so.
This is the second case of copyright infringement by writers against artificial intelligence that courts have partially dismissed. Last November, Judge Vince Chhabria also dismissed portions of Sarah Silverman’s allegations against Meta.
Despite the authors’ claims, OpenAI states that results generated by ChatGPT are not similar enough to copyrighted materials to violate the holders’ rights. Judge Martinez-Olguin echoed a similar sentiment and stated the authors “fail to explain what the outputs entail or allege that any particular output is substantially similar — or similar at all — to their books.”
The courts have yet to state whether these companies’ unauthorized use of copyrighted material as training fodder violated copyright laws. The tech giants have refuted such claims and insist the fair use doctrine protects their usage. The companies also raised concerns about how such lawsuits may threaten the overall growth of the AI industry.
Larger Battle Against AI
Sarah Silverman and Paul Tremblay’s battles against AI are just the tip of the iceberg. Groups of creators, from visual artists to musicians and writers, launched a multipronged legal war against multiple major tech companies. These creators share the same concern — tech giants’ unauthorized use of their material to train their systems is copyright infringement on a massive scale. Authors like George R.R. Martin, John Grisham, and Jodi Picoult also joined the fray last year with lawsuits of their own.
Microsoft Roped Into the Legal Battle
In November last year, nonfiction writer Julian Sancton filed suit against Microsoft in New York federal court. Like the other prolific lawsuits, Sancton claims that OpenAI engaged in “rampant theft” of copyrighted material to train its AI system.
Additionally, the claim alleges that Microsoft’s active and critical support enabled such violations of artists’ rights. It notes that Microsoft played a “key role” and provided “critical assistance,” addressing the utilization of tech like Microsoft Azure while knowing about OpenAI’s manipulation of licensed material to develop the system. According to Sancton’s claims, Microsoft has been “deeply involved in the training, development, and commercialization” of these systems. They provided the developers with Azure — a specialized computing system that played a key role in the development of the model.
An Artist’s Right To Pursue Lawsuits
While the situation may seem grim for creatives, they have secured small victories. Last year, a group of visual artists filed a lawsuit against StabilityAI, Midjourney, and others, asserting that the companies had violated copyrights by using the artists’ works to train their AI without their knowledge or proper compensation.
The companies retaliated, saying that the suit targeted their right to free speech by hindering the creation of art. They asked for an early dismissal of the suit, claiming they were entitled to First Amendment defense under the California statute. Judge William Orrick dismissed their claims of being entitled to First Amendment protection. He said the case is in the public interest.
The Case of Books3
Last year, journalist Alex Reisner revealed the existence of Books3 — a once-searchable book data set. In his report for The Atlantic, he revealed this data set has over 170,000 books, including ones by Zadie Smith, Margaret Atwood, Stephen King, and Michael Pollan. Companies like Bloomberg and Meta used this data set to train their AI models.
Books3 echoes the spirit of past data sets, including ones OpenAI trained its model om, like Books1 and Books2. Though independent AI researcher Shawn Presser created Books3 for scientific purposes, the existence of this shadow library reveals the sheer breadth of the problem and shows how AI firms often ignore artists’ rights in favor of advancing generative AI.
