US authors sue OpenAI and Microsoft over training data

Background

Copyright law gives authors control over how their books are copied and distributed. AI systems that generate text are trained on large collections of written material, and the use of books in that process has become a legal dispute. In these cases, the authors argued that the companies copied their books in bulk from the internet without permission to train and develop chatbots^[1]. The disputes were filed as part of a wave of copyright litigation against Microsoft and OpenAI in late 2023 and early 2024^[2].

What happened

On 20 December 2023, eleven US authors, including Taylor Branch and Stacy Schiff, sued OpenAI and Microsoft in the Manhattan federal district court^[1]. They said the companies copied their books in bulk from the internet without permission to train and develop chatbots^[1]. They sought unspecified damages and an order stopping the companies from infringing their copyrights^[1].

Key details

On 5 January 2024, authors Nicholas Basbanes and Nicholas Gage filed a class action in the same court against Microsoft and OpenAI^[2]. The complaint accused the companies of “systematic theft on a mass scale” of copyrighted works^[2]. The suits argued that the companies used information from the authors’ books to train AI products including ChatGPT^[2]. A class action is a procedure in which one or more plaintiffs ask to represent a larger group of people with similar claims; here, the filing was brought by two named authors^[2]. The earlier suit was brought by eleven named authors^[1]. Both matters concerned the same basic allegation: that books were copied for AI training without authorization^[1][2].

Why it matters

The author suits were part of a wave of copyright litigation against Microsoft and OpenAI in late 2023 and early 2024^[2]. They put the question of whether using books to train chatbots requires permission before the courts. That question matters because AI products including ChatGPT are built from training data, and authors say their books were part of that data without their consent^[2]. In the first suit, the authors sought money damages for the alleged copying and an injunction to stop ongoing infringement^[1]. Such claims ask a court to weigh copyright protection for authors against the development and operation of AI products that use large text collections^[1][2].