Coinspeaker
Authors File Copyright Lawsuit against OpenAI for Using Their Books to Train ChatGPT without Consent
Two authors have filed a class-action lawsuit against ChatGPT creator OpenAI for allegedly infringing on their copyright, and using their books to train the artificial intelligence (AI) chatbot without obtaining consent. According to the suit, ChatGPT can produce “very accurate summaries” of their books, which is “only possible” if OpenAI used their books to train the chatbot.
Authors Paul Tremblay and Mona Awad filed the lawsuit in a San Francisco federal court. Their lawsuit states that OpenAI used a lot of data protected by copyright law to train ChatGPT, including their books. The authors state that OpenAI exploits published material and that they want financial compensation on behalf of authors in the US whose publications OpenAI “unfairly” benefit from.
Both authors have rights to and have published several books, including Tremblay’s “The Cabin at the End of the World”, as well as Awad’s “13 Ways of Looking at a Fat Girl” and “Bunny”. Included in the complaint are a few “exhibits” of ChatGPT’s responses containing summaries of these books, which include some wrong information. However, the complaint states that ChatGPT “retains knowledge of particular works in the training dataset” since the summaries are largely accurate.
University of Sussex intellectual property law reader Andres Guadamuz confirms in a conversation with The Guardian that this is the first copyright lawsuit against OpenAI’s ChatGPT. Guadamuz believes that the lawsuit could clarify what constitutes legal actions in the general AI space, especially with generative text.
Is There a Legitimate Copyright Case against OpenAI?
According to the filing, OpenAI owes plaintiffs a duty of care to protect obligations, customs, and practices, and reasonably control the information it holds. The lawsuit then states:
“Defendants breached their duties by negligently, careless, and recklessly collecting, maintaining and controlling Plaintiffs’ and Class members’ Infringed Works and engineering, designing, maintaining and controlling systems – including ChatGPT – which are trained on Plaintiff’s and Class members’ Infringed Works without their authorization.”
However, Guadamuz suggests that the plaintiffs may not have much of a case. According to him, proving that Tremblay and Awad were financially exploited by ChatGPT’s use of their publications could be tricky even if there is proof ChatGPT appropriated their works. Guadamuz believes that ChatGPT could work “exactly the same” even if these books were not used for its training. This is because much of the chatbot’s training relies on public information on the internet, which may include conversations about books.
Newcastle University professor of law, innovation and society, Lilian Edwards, believes the court’s interpretation may be the most important. According to Edwards, a resolution of the copyright lawsuit against Open would depend on the court’s understanding of OpenAI’s use of copyrighted material. Edwards says plaintiffs may win or lose if the courts determine OpenAI’s application of the books is “fair use” or “simple unauthorized copying”.
Interestingly, she states that the court’s interpretation would be different in the UK since the country does not allow the “fair use” justification in the same way the US might.
Authors File Copyright Lawsuit against OpenAI for Using Their Books to Train ChatGPT without Consent
