Meta CEO Mark Zuckerberg will be deposed as part of a lawsuit filed by authors concerning the company’s use of artificial intelligence technology.
 |
| Mark Zuckerberg Alex Wong/Getty Images |
On Tuesday, U.S. District Judge Thomas Hixson rejected Meta’s attempt to block Zuckerberg’s deposition, citing evidence that he is the “principal decision maker” for the company’s AI initiatives.
Authors Sarah Silverman, Richard Kadrey, and Christopher Golden filed a proposed class action last year in California federal court. They accused Meta of copyright infringement, alleging that the company illegally downloaded their books from shadow library websites and used them without permission or payment to train its AI system.
This case is one of several legal challenges from creators regarding how large language models are trained, and it could help establish guidelines for developing such technologies.
In his ruling, Judge Hixson stated that Zuckerberg is the “policy setter” for Meta’s generative AI branch and plays a direct role in the development of the AI systems in question.
The authors provided “evidence of his specific involvement in the company’s AI initiatives,” along with his “direct supervision of Meta’s AI products,” Hixson wrote.
Meta argued that Zuckerberg lacks unique knowledge about the company’s operations that would justify his deposition, claiming such information could be obtained from other employees or executives. They emphasized that the central issue in the case is fair use, which allows copyrighted material to be used without a license or compensation when creating secondary works.
Meta’s lawyers noted that “fair use will primarily depend on the transformative nature of the AI models and the alleged impact on the market for Plaintiffs’ books,” and that the authors “do not need numerous depositions, let alone Zuckerberg’s deposition, to establish or refute these elements.”
In response, the authors argued that Zuckerberg had personally issued directives guiding the development and sale of Meta’s AI products. They referenced a New York Times article reporting that he “immediately pushed to match and exceed ChatGPT, calling executives and engineers at all hours to push the development of a rival chatbot.”
The court ultimately agreed that the authors had “made a solid case that this deposition is worth taking.”
Meta has not disclosed the specific origins of the books used in its data set to train LLaMA. While the company claims the data came from publicly available sources, it has not provided further details.
OpenAI is facing a similar copyright infringement lawsuit from writers. On Tuesday, both parties reached an agreement on protocols to inspect the company’s training data to determine whether copyrighted works were used in developing its technology.
0 Comments