OpenAI, the organization behind the ChatGPT platform, finds itself amidst a series of lawsuits centered on allegations of copyright infringement. For context, ChatGPT operates as a free-to-use generative AI model trained on vast amounts of text data from the internet. This extensive training allows it to generate human-like text based on user prompt inputs. Since its public release in late 2022, its adoption has skyrocketed, with the platform now boasting over 1.5 billion monthly visits.

That said, the sources of its training data, potentially including copyrighted works, have become a point of contention. As of now, there are just a few filed lawsuits that highlight the intricate relationship between AI training data and copyrighted content.

The case of Authors v. OpenAI serves as a prime example. Here, the plaintiffs (a consortium of authors) assert that OpenAI’s models, in their training phase, consumed copyrighted books, articles, and other literary works that, despite being publicly accessible, remain under copyright protection. Their contention essentially is that such training equips the models to either mirror original content or produce derivative works. In fact, the plaintiffs have described instances where the models exhibited this behavior. They argue that this erodes their creations’ intrinsic value and robs them of due royalties. The lawsuit, grounded on theories of direct and vicarious copyright infringement, negligence, and state unfair competition law violations, seeks monetary damages and an injunction against OpenAI’s continued use or distribution of models trained on copyrighted content.

Parallel lawsuits, like Sarah Silverman et al. v. OpenAI and Paul Tremblay & Mona Awad v. OpenAI, echo similar complaints. Media giant, The New York Times, is also contemplating legal measures against OpenAI. In a broader context, other AI entities are facing litigation for generative AI models related to visuals (not just text generation), such as in Getty Images v. Stability AI (in the High Court of Justice in London).

While generative AI, in its foundational concepts, has been around for several decades its rapid development and accessibility to the general public are recent phenomena. As a result, the legal landscape is still in its infancy, and legal doctrine sparse. While rooted in longstanding law, the cases at hand are trailblazers in their own right, set to define the legal boundaries of the generative AI frontier.