Writers sue Meta and OpenAI for Infringement, Questioning the Legality of AI Training Data
Renowned writers filed lawsuits against Meta and OpenAI for infringement, and the result could set a legal standard for AI training data, or even the destruction of relevant AI systems trained on copyrighted content.
A new legal storm is brewing in the landscape where entertainment and technology intersect. Renowned authors and screenwriters, led by Michael Chabon, are escalating their efforts against Meta and OpenAI, alleging copyright infringement. These writers claim the companies are using their copyrighted works to train AI systems, thus producing potentially infringing derivative content, and thus, what the two tech giants are facing are court challenges that will decide the legality of how their popular and widely-used AI tools were trained.
The methodology employed by relevant tech companies have been in question, and is particularly the heart of this case. Allegedly, vast amounts of literary content are being harvested from the web, notably from controversial datasets like BookCorpus. BookCorpus was created in 2015 by a team of AI researchers by Google and Samsung and has faced criticism for incorporating numerous copyrighted works without author consent or compensation. OpenAI revealed the use of over 7,000 novels on BookCorpus for training the first iteration of their large language model, GPT-1. The implications are far-reaching, especially when AI models like ChatGPT can provide in-depth analyses and summaries of literary works, hinting at their foundational training on these copyrighted texts.
This isn't the first time tech companies have faced such legal challenges. In 2005, Google's ambitious project to digitize millions of books met with resistance from the Authors Guild. The tech giant's defense, however, rooted in the idea of 'fair use,' and prevailed as the court ruled that only snippets, not entire books, were accessible to users. A recent Supreme Court decision in the Andy Warhol Foundation for the Visual Arts v. Goldsmith case, on the other hand, headed in a different direction. The court highlighted the commercial implications of derivative works, emphasizing their similarity to original content as a potential infringement and rejecting the defense of fair use.
Beyond the legal ramifications, there's a broader concern at play. AI's increasing role in content generation raises fundamental questions about the future of creativity and originality in the entertainment industry. Notable authors such as Stephen Chbosky, Emma Donoghue, and Gillian Flynn all have adapted their novels to screenplays. It is worth noting that if AI can produce screenplays in the style of a famous book or author and gain a more significant role in the space, it can easily undermine the market prospects for writers.
Ed Klaris, an IP lawyer, highlights that the courts' approach is likely to revolve around the nature of use. If AI companies begin producing derivative works that compete directly with the original, the legal landscape may start tilting in favor of the authors. Klaris speculates that if the courts favor creators in their analysis of fair use, it might force AI companies to develop a licensing framework. Such a move would recognize and potentially remunerate authors for their foundational work.
The writers seek a court order that would require the companies to destroy AI systems that were trained on copyright-protected works. Neither OpenAI nor Meta has made a comment on the case, and all the industry can do right now is watch the legal battle unfold.