OpenAI is trying to dismiss various claims in two legal actions launched by authors and comedians, who sued the machine-learning super-lab for scraping their books to train ChatGPT without explicit permission. 

In June, novelists Paul Tremblay and Mona Awad filed a lawsuit against OpenAI in the US over its harvesting of prose. Days later, comedian Sarah Silverman and novelists Christopher Golden and Richard Kadrey also sued the biz. Both parties accused the ChatGPT maker of copyright infringement for ingesting their work to train the chatbot.

OpenAI’s lawyers have hit back, asking a federal court in San Francisco to dismiss five out six claims brought by Tremblay and Awad, those being: vicarious infringement, violation of the Digital Millennium Copyright Act, unfair competition, negligence, and unjust enrichment. It is, however, prepared to fight the first claim of direct copyright infringement in hope of winning it to make a point to any other creative person considering pursuing it.

OpenAI made the same argument against Silverman et al’s federal lawsuit, too, for what it’s worth.

“This motion does not seek dismissal of Count I, for direct copyright infringement, which OpenAI will seek to resolve as a matter of law at a later stage of the case,” the lab’s legal eagles argued in court filings [PDF]. The lawyers believe most of the plaintiffs’ accusations do not describe any actual law-breaking by OpenAI, and rest allegedly involve the biz directly infringing on the writers’ intellectual property – allegations the organization denies. 

Lawyers for the AI biz argued OpenAI has not violated copyright laws at all and that ChatGPT is protected under fair use. US copyright law states that “transformative uses” of work – where the original source is repurposed – is considered fair use. Although the large language model ingested the authors and comedian’s books, it transformed their text for different applications. That’s the argument, anyway. 

“According to the complaints, every single ChatGPT output – from a simple response to a question (e.g., ‘Yes’), to the name of the president of the United States, to a paragraph describing the plot, themes, and significance of Homer’s The Iliad – is necessarily an infringing ‘derivative work’ of plaintiffs’ books,” they wrote in the court docs.

“Worse still, each of those outputs would simultaneously be an infringing derivative of each of the millions of other individual works contained in the training corpus – regardless of whether there are any similarities between the output and the training works. That is not how copyright law works,” they concluded.

AI and copyright is a contentious legal gray area. Similar lawsuits have been filed by visual artists, who claim companies like Stability AI have trained text-to-image models on their artwork. Although the US Copyright Office has declared that works which are “not the product of human authorship” cannot be protected, officials are unsure about other issues.

The office issued a request for public comment [PDF] this week on copyright law and policy issues raised by AI.

The Register has asked the plaintiffs’ lawyers for comment. ®

Read More