Judge refuses to block California AI training data disclosure law

In short: A federal judge refused to pause California’s AB 2013, so a new rule requiring AI companies to disclose training data sources can move forward.

What happened

A federal judge in California denied Elon Musk’s AI company, xAI, a preliminary injunction, which is an early court order that would have temporarily stopped a law while the lawsuit continues.

The law is California Assembly Bill 2013, also called the Generative Artificial Intelligence: Training Data Transparency Act. It was signed in September 2024 and took effect on January 1, 2026. It requires companies that make “generative” AI (systems that can create text or images, like a very fast content generator) to publicly describe the data sources used to train their models, including text, images, and code.

xAI sued California Attorney General Rob Bonta in late December 2025. xAI said the lists of training data are trade secrets, meaning valuable information a company keeps private, like a recipe. It also argued the law would force it to speak in public against its will and that it could unfairly take company property by making it hand over sensitive details.

The judge rejected those arguments and allowed the law to proceed. Reports also noted that the state did not give a clear timeline for enforcement during the hearing, which can make it harder for a court to treat the situation as an emergency.

Why it matters

This decision supports the idea that the public can ask basic questions about what AI systems learned from. For regular people, it can help explain why an AI tool might copy personal data, repeat harmful stereotypes, or create unsafe content, because training data is the material the system learned from (like what a student reads before a test).

Source: Arstechnica