XAI v Bonta

Case Information

Case name	XAI v. Bonta
Court	U.S. District Court for the Central District of California
Docket	2:25-cv-12295-JGB-SSC; appeal 26-1591
Filed	December 29, 2025
Judge(s)	Jesus G. Bernal (district court)
Plaintiffs	X.AI LLC
Defendants	Rob Bonta, California Attorney General, in his official capacity
Case type	Civil constitutional action and interlocutory appeal
Claims / issues	First Amendment compelled speech, takings and due-process challenges to AB 2013 training-data disclosures
Status	March 4, 2026: preliminary injunction denied. The appeal is identified in the filed appellate brief; later appellate disposition not verified.

X.AI LLC v. Rob Bonta (Case No. 2:25-cv-12295-JGB-SSCx) is a constitutional challenge by Elon Musk's xAI Corporation against California Assembly Bill 2013, which requires generative AI developers to disclose high-level summaries of their training datasets. The U.S. District Court for the Central District of California denied xAI's motion for a preliminary injunction on March 4, 2026, marking the first federal court ruling upholding a state AI transparency law.

Background

California Assembly Bill 2013, signed into law in 2024 and effective January 1, 2026, requires developers of generative AI systems to publicly disclose a "high-level summary" of the datasets used to train their models, including the number of data points, sources, whether the data includes copyrighted or protected content, and how the data was curated. The law is the first of its kind in the United States to mandate AI training data transparency.

xAI's Arguments

xAI sought a preliminary injunction to block enforcement of AB 2013, raising three constitutional challenges:

• First Amendment (Compelled Speech): xAI argued that AB 2013 forces AI developers to disclose proprietary information about their training datasets, constituting compelled speech. xAI characterized its dataset information as commercial speech subject to Central Hudson intermediate scrutiny, arguing the law cannot survive such review.

• Fifth Amendment Takings Clause: xAI claimed that mandatory disclosure of training data details destroys trade secrets in curated datasets, their sizes, sources, processing methods, and inclusion of protected data, constituting a regulatory taking without just compensation.

• Fourteenth Amendment Due Process (Vagueness): xAI argued that terms like "high-level summary" and "dataset" are unconstitutionally vague, failing to give fair notice of what disclosure is required.

Court Ruling

On March 4, 2026, Judge Bernal denied xAI's motion for a preliminary injunction after oral argument on February 23, 2026, finding xAI unlikely to succeed on the merits of any of its constitutional claims.

Standing

The court found xAI had standing, as it demonstrated potential harm from enforcement and the Attorney General did not disavow enforcement.

Takings Clause

The court rejected xAI's takings claim, finding xAI's general allegations insufficient — xAI provided no proof of unique datasets, specific sizes, or proprietary processes that would be destroyed by disclosure. The court noted that alternatives like inverse condemnation suits exist for any actual takings that might later occur.

First Amendment

The court held that AB 2013 survives Central Hudson intermediate scrutiny: it advances substantial governmental interests in enabling consumers to evaluate AI systems through dataset information, and the means employed are no more extensive than necessary to achieve those interests.

Vagueness

The court rejected the vagueness challenge, finding that terms like "high-level summary" are clarified by the statute's enumerated disclosure requirements (e.g., sources, data points, presence of protected content), and that xAI itself demonstrated sufficient understanding of the requirements to challenge them.

Appeal

By July 24, 2026, xAI's challenge was before the U.S. Court of Appeals for the Ninth Circuit on appeal from the denial of preliminary injunctive relief. Open Markets Institute joined an amicus brief authored by the AI Coalition of Data Integrity urging the Ninth Circuit to affirm the district court's ruling. The brief argues that xAI failed to show that AB 2013's required training-data disclosures violate the First Amendment.

Significance

This is the first federal court ruling on the constitutionality of a state AI transparency law. The ruling has significant implications for:

• Federal preemption debates: The ruling undermines arguments that state AI transparency laws are constitutionally problematic, strengthening the case for state-level AI regulation.
• Trade secret protections: The court's narrow reading of takings claims signals that AI companies must provide specific evidence of trade secret harm, not generalized allegations.
• Compelled speech doctrine: The ruling suggests that disclosure requirements for AI training data survive intermediate scrutiny under the First Amendment.

The preliminary-injunction ruling is now on appeal in the Ninth Circuit. Regardless of the ultimate outcome, Judge Bernal's detailed order provides a roadmap for defending state AI transparency laws against constitutional challenges.

See Also

• xAI v. Bonta: Court Upholds California AI Transparency Law
• California AB 2713 Adjusts AI Transparency Act
• Cooley State AI Laws Update
• Cases — Active AI litigation tracker

• U.S. catastrophic-risk policy: comparative context

References