
Free Speech Arguments – Are AI “Trade Secrets” Protected by the First Amendment? (X.AI LLC v. Rob Bonta)
Episode 59: X.AI LLC v. Rob Bonta
X.AI LLC v. Rob Bonta, argued before Judges Johnnie B. Rawlinson, Milan D. Smith, Jr., and Danielle J. Forrest of the United States Court of Appeals for the Ninth Circuit on November 18, 2026. Argued by James Y. Xi (on behalf of X.AI LLC) and Joseph Meeker (on behalf of the state of California).
Case Summary, from the Brief for Appellant:
Like many companies in the burgeoning artificial intelligence (“AI”) industry, X.AI LLC (“xAI”) develops generative artificial intelligence models, or gen-AI models for short. Gen-AI models are software programs that simulate human intelligence to create original content and perform a wide range of tasks—everything from answering trivia questions and summarizing documents to generating images and videos. To train its models, xAI compiles “datasets”—collections of information derived from different sources. xAI plugs those datasets into its AI models, which users can use to generate new content. Those models, however, are only as good as their training, which is only as good as the data on which they are trained. Unsurprisingly then, xAI (like its competitors) dedicates substantial resources to identifying and developing high-quality training data, particularly from sources its competitors are not using. And unsurprisingly, xAI (like its competitors) treats such information as proprietary trade secrets. As companies race to build the next successful gen-AI product, the competitive advantage provided by data quality and training efficiency has become increasingly important…
Though companies in the industry fiercely guard such information from their competitors, California Assembly Bill 2013 (“A.B.2013”) requires AI companies to disclose that information to the public. But just as the government could not require, e.g., Coca-Cola to reveal its formula for Coke to the public, California cannot compel AI companies to reveal proprietary information about the datasets they use to train their gen-AI models. In fact, A.B.2013 is unconstitutional thrice over: It violates the First Amendment by compelling xAI to disseminate confidential information about its generative AI models. It violates the Takings Clause by forcing xAI to disclose its economically valuable and fiercely protected trade-secret property rights without any promise of compensation. And it is unconstitutionally vague because it fails to provide regulated parties with sufficient clarity about which datasets it covers and how much detail about that proprietary information they must provide.
Statement of the Issues from the Brief for Appellant:
- Whether xAI is likely to succeed in showing that A.B.2013 violates xAI’s First Amendment rights.
- Whether xAI is likely to succeed in showing that A.B.2013 effects an uncompensated taking in violation of the Takings Clause.
- Whether xAI is likely to succeed in showing that A.B.2013 is unconstitutionally vague.
Resources:
The Institute for Free Speech promotes and defends the political speech rights to freely speak, assemble, publish, and petition the government guaranteed by the First Amendment. If you’re enjoying the Free Speech Arguments podcast, please subscribe and leave a review on your preferred podcast platform.
Altri episodi di "Free Speech Arguments"



Non perdere nemmeno un episodio di “Free Speech Arguments”. Iscriviti all'app gratuita GetPodcast.








