Judge Rejects xAI Bid to Block California AI Training Disclosure Law
Mar 9th 2026
A federal judge refused xAI's attempt to block California's AB 2013, which requires AI developers to disclose details about training data and raises new legal pressure over the use of copyrighted material in model training.
- U.S. District Judge Jesus Bernal denied xAI's request to block AB 2013, leaving the law in force.
- AB 2013 requires generative AI developers available in California to publish a 'high-level' summary of training datasets, including sources, copyright status, licensing, and first-use dates.
- The statute does not define 'high-level summary,' creating a vagueness argument that xAI may continue to press in court.
- Ninth Circuit precedent in MAI and Sega indicates that intermediate or temporary copying during training can count as reproduction under the Copyright Act.
- Many frontier AI developers scraped large volumes of internet content, so their training sets likely include copyrighted material and fair use remains an untested defense at scale.
- AB 2013 has no direct penalties but could enable California enforcement actions, aid private plaintiffs in discovery, and change the litigation landscape for ongoing copyright suits.