Today, LASST and a coalition of 28 spanning civil society organizations, child safety advocates, AI law and policy groups, and First Amendment scholars filed an amicus brief in the Ninth Circuit defending California’s AI Training Data Transparency Act (AB 2013) against a constitutional challenge brought by xAI.
What xAI v. Bonta is about
AB 2013 requires developers of generative AI systems to publish a high-level summary of the data used to train their models: what kind of data it is, and whether it includes personal information, copyrighted material, or synthetic data.
What it doesn’t require is that companies hand over their datasets, disclose code, or reveal proprietary algorithms. It’s a transparency law, not a data-dump mandate.
xAI, owned by Elon Musk and developer of the Grok chatbot, argues that even this high-level disclosure violates its First Amendment rights. A federal judge in the Central District of California already rejected that argument once. xAI is now asking the Ninth Circuit to reverse that decision.
AB 2013 is a run-of-the-mill disclosure law. Courts have upheld disclosure requirements like this one for decades: ingredient labels; country-of-origin markings; drug-interaction warnings. AB 2013 fits comfortably within that tradition, rooted in consumer protections. It’s an ingredient label for AI products.
Why it matters
Transparency into training data isn’t an abstract interest. We’ve already seen real-world risks emerge in the AI products we are using right now. Parents should have the right to know whether the chatbot on their kid’s phone was trained on harmful material. Financial advisers with fiduciary duties need to know what’s inside the tools they use professionally. Businesses in regulated industries face liability exposure from AI systems that may have been trained on data that violates copyright or privacy law.
AB 2013 gives consumers the information they need to make informed decisions in a market that badly needs more of it.
Transparency into how AI companies develop and deploy their models is critical to public safety and is in the public interest, especially as the technology becomes more capable and more widely adopted. The First Amendment should not be distorted to shield AI companies from basic product disclosure requirements that apply to every other industry.
We’re not the only ones
The wide array of organizations and scholars signing onto today’s filing is in itself an answer to xAI’s claim that consumers have no real interest in training data transparency.
The bigger stakes
This will be the first appellate decision on the constitutionality of state AI training-data transparency laws. How things turn out will shape how courts and regulators think about disclosure-based AI governance going forward, with direct consequences for other transparency laws already on the books, including California’s SB 53, New York’s RAISE Act, and Illinois’ SB 315.
These disclosure requirements are about as light-touch as AI regulation gets. They don’t tell developers what to build or how to build it. They simply require developers to tell the public, at a high level, what went into creating the products they’re already selling. The Ninth Circuit’s own precedent supports that approach, and we’re glad to stand with this coalition in defending it.
Our 29 coalition members include:
Coalition members include Alliance for Secure AI, Americans for Responsible Innovation (ARI), California Initiative for Technology and Democracy (CITED), Center for Humane Technology, Cybersafety Research Center, Digital Childhood Institute, Electronic Privacy Information Center (EPIC), Encode AI, Fairplay, David Atkinson of Future of Life Institute (signing in his personal capacity), Heat Initiative, Legal Advocates for Safe Science & Technology (LASST), NAVA, National Center on Sexual Exploitation (NCOSE), Secure AI Project, Surveillance Technology Oversight Project (STOP), Tech Equity, Tech Justice Law (TJL), Tech Oversight Project, and Transparency Coalition.
They are joined by Stuart Russell OBE FRS, Distinguished Professor of Computer Science at UC Berkeley; Lawrence Lessig, Roy L. Furman Professor of Law at Harvard Law School and founder of Creative Commons; and legal scholars Margot Kaminski (University of Colorado Law), Helen Norton (University of Colorado Law), Olivier Sylvain (Fordham Law), Kyle Langvardt (University of Nebraska College of Law), Peter Ormerod (Villanova Law), Vivek Krishnamurthy (University of Colorado), and Zephyr Teachout (Fordham Law).
The full brief explaining the stakes for consumers is available here.