xAI LLC v. Bonta - AI Training Data Disclosure Amicus Brief

Status: Filed; Legal Advocates for Safe Science & Technology (LASST), joined by a coalition of 28 scholars and civil society groups, filed an amicus brief urging the Ninth Circuit to affirm the district court’s denial of xAI’s motion for a preliminary injunction and to uphold California’s AI training data disclosure law.

Court: U.S. Court of Appeals for the Ninth Circuit

Key issues: AI Transparency; Commercial Speech; First Amendment; Consumer Protection

xAI is challenging California’s Assembly Bill 2013, which requires developers of generative AI systems to post a “high-level summary” of the datasets used to train models made publicly available to Californians. The law does not require developers to produce the underlying datasets, disclose proprietary algorithms, or reveal source code. It asks only what kinds of data went into a model and where that data came from. xAI sued California’s Attorney General, Rob Bonta, arguing the disclosure mandate violates its First Amendment rights; the district court denied xAI’s request for a preliminary injunction, and xAI appealed to the Ninth Circuit.

LASST and its amici argue that AB 2013 is a routine, constitutional disclosure law:

• AB 2013’s disclosures are commercial speech, i.e., factual, product-specific information akin to ingredient labels, nutrition facts, or country-of-origin labeling, not speech that conveys any independent expressive or political viewpoint.

• The deferential Zauderer standard for commercial speech applies, because the disclosures are purely factual and uncontroversial, and under that standard, the law easily survives rational basis review.

• California has substantial interests in requiring the disclosures, including correcting severe information asymmetries in a fast-moving and highly concentrated market, protecting consumers, safeguarding public health, preventing the sexual exploitation of children (given documented instances of CSAM in AI training datasets), and protecting individual privacy.

• The disclosures are neither unduly burdensome nor unjustified: xAI has not identified any concrete burden, remains free to supplement its disclosures with additional context, and has already complied with AB 2013 without apparent harm to its business.

The Ninth Circuit’s resolution of this appeal will be one of the first appellate rulings to address the constitutionality of AI-specific transparency mandates, and amici argue its reasoning, not just its outcome, will shape how courts and legislatures approach AI disclosure requirements going forward. Amici frame disclosure laws as the “lightest touch” mode of AI regulation available to policymakers, arguing that requiring companies to share basic, factual information about what their models were trained on is a modest and well-established tool for addressing the information gap between AI developers and the public.