NewsAI regulationCaliforniaTransparency

xAI argues California's AI training data law is unconstitutional

Before the Ninth Circuit, xAI is asking for California's AB 2013 to be blocked. The law requires AI developers to publish summaries of their training data. No ruling date has been set.

AB 2013 in force since January 1

xAI argues California's AI training data law is unconstitutional

The lawsuit

On Wednesday, November 18, xAI asked a three-judge panel of the US Court of Appeals for the Ninth Circuit to block California's AB 2013. The law on AI training data transparency is unconstitutional, the company argues, and it exposes trade secrets. The case is X.AI LLC v. Bonta, No. 26-1591, heard in Pasadena by Judges Rawlinson, Smith and Forrest. According to MLex, the panel gave little sign of how it will rule.

What the law requires

Governor Newsom signed AB 2013 in September 2024. The law took effect on January 1. Developers of generative AI systems must post on their own websites a summary of the datasets used to train them. The law asks for high-level information, not the datasets themselves. California relies heavily on this point.

xAI bases its complaint on two claims: the law forces the company to disclose trade secrets it values in the billions of dollars, and it compels speech in violation of the First Amendment, because the state dictates what a company must say about its own product.

Where the case stands

xAI lost the first round. A federal judge in California denied its request for a preliminary injunction earlier this year, as Law360 reported. Judge Jesus Bernal accepted that training data could in principle be a trade secret, but said xAI had not shown its datasets or data-processing methods were more sensitive than those of rivals. He also called xAI's claim that no consumer could learn anything useful from a dataset summary a stretch.

The denial stays in place unless the appeals court decides otherwise. The law is therefore in force while the case runs.

No decision in sight

No ruling date has been set. Amicus briefs were due in July, and the Ninth Circuit has not said when it will decide. According to MLex, AI researchers have filed a brief urging the court to uphold the law.

Assessment

The ruling could become one of the first appellate decisions on whether AI-specific transparency mandates survive the First Amendment. Rulings from this court carry weight across the western states, and other legislatures drafting similar bills will read the opinion closely. A win for xAI could serve other labs as a template for challenging comparable rules. A loss would classify a public training data summary as an ordinary business disclosure.

For developers shipping models to California users, the practical point is that OpenAI, Anthropic and Google have not joined the challenge, and xAI is the lone company plaintiff. Anyone operating such a model should plan for the disclosure obligation regardless of how the case ends. Legal commentary also warns that a disclosure that looks harmless on a compliance checklist can become evidence in a copyright or privacy suit.

Sources

Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.

Share
← All articles

All analyses are based on i6eal's own measurements or on clearly labelled sources. Figures are snapshots and may change; corrections are disclosed transparently.