xAI Loses Bid to Block California AB 2013 AI Training Data Disclosure Law

A US federal judge denied xAI's motion for a preliminary injunction to block California's AB 2013, which requires AI developers whose models are accessible in California to publicly disclose information about the datasets used to train them, including sources, collection dates, copyright status, and use of personal data. Judge Jesus Bernal ruled that xAI failed to demonstrate its datasets were sufficiently unique to qualify as trade secrets under the Fifth Amendment, and rejected xAI's argument that the public has no interest in AI training data disclosures. xAI must now comply with the law during its ongoing litigation, potentially revealing training data details that Musk feared could benefit rivals like OpenAI.

Key Takeaways

• California AB 2013 (in effect January 2026) requires AI developers to disclose dataset sources, collection dates, copyright status, personal data use, and synthetic data percentage for models accessible in California
• Judge Bernal denied xAI's preliminary injunction; ruled xAI's Fifth Amendment (trade secret) and First Amendment claims insufficient — xAI could not show its datasets or cleaning methods were meaningfully unique
• xAI must now comply with AB 2013 during continued litigation; non-compliance risk: OpenAI or other rivals gaining visibility into xAI's training data sourcing strategy

Original source: Ars Technica