Title: Authors File Lawsuit Against OpenAI for Training ChatGPT with Their Books: Will They Succeed?
Authors Mona Awad and Paul Tremblay have recently filed a lawsuit against OpenAI, alleging that their books were used without their consent to train ChatGPT, an artificial intelligence program developed by OpenAI. This is the first copyright lawsuit concerning ChatGPT, and it raises the question of whether or not the authors have a valid case.
In this scenario, instead of a human reading a book and providing a written response to a question about it, OpenAI is accused of copying the books into its internal database and training ChatGPT to produce accurate summaries of the works when prompted.
The lawsuit claims that OpenAI utilizes shadow libraries that illegally distribute copyrighted works via torrent systems. The basis of the lawsuit centers around a 2020 paper by OpenAI, which revealed that 15% of their training dataset came from two internet-based books corpora. However, the lawsuit does not specify which parts of Awad and Tremblay’s novels have been unlawfully copied and reproduced in the summaries.
For the lawsuit to succeed, the authors must prove that OpenAI most likely copied their works and demonstrate the likelihood of economic loss. However, copyright protection does not extend to ideas; it is limited to the written expression. Although copying the books into a database may constitute infringement, it alone may not cause significant harm to the authors’ economic interests.
The main concern is that OpenAI’s capabilities resemble those of human authors. This raises the question of how Australian law would handle a similar claim. Would fair dealing laws protect the development of technology, or would they side with the authors?
In the United States, fair use doctrine exists in copyright laws to balance new technologies with established copyright interests. Australia, on the other hand, relies on the doctrine of fair dealing, which is similar but less flexible. While the Australian Copyright Act contains provisions on time-shifting and fair dealing for parody, it does not incorporate fair use.
Australian law, like the United States, protects tangible expression but not ideas. This means that people must be free to use ideas in subsequent works. When applied to large-language models like OpenAI, the same logic should apply.
While the OpenAI litigation may not succeed, it signals the beginning of a significant shift in copyright law due to advancements in artificial intelligence. Australia will need to engage in a policy debate on how to navigate the potential conflicts between evolving technologies and the livelihoods of authors.
The outcome of this lawsuit remains uncertain, but it highlights the need for a comprehensive legal framework that strikes a balance between technological advancements and authors’ rights. As the first lawsuit of its kind, it sets the stage for future challenges in an AI-driven landscape.
