X.AI LLC v. Bonta

MOTION to Extend Time to File Brief — Attachment 18

Issue: In *X.AI LLC v. Bonta*, California Attorney General Rob Bonta argues that the appeal turns on two unsettled legal questions: whether AI developers hold cognizable trade secrets in the training data used to build generative AI models, and what level of First Amendment scrutiny applies to a California statute requiring disclosure of that data. Neither question has established Ninth Circuit precedent, and the answers could determine the constitutional limits of AI-specific transparency regulation more broadly.

At the pre-briefing stage of this Ninth Circuit appeal, Appellee Rob Bonta filed an unopposed motion on May 19, 2026, seeking a 30-day extension under Ninth Circuit Rule 31-2.2(b) to file his answering brief in response to X.AI LLC's opening brief. The motion seeks to move the answering brief deadline from June 15 to July 15, 2026. Bonta's office justifies the extension on two grounds: the complexity and novelty of the underlying AI regulation issues, which require substantial research, and the need for multiple rounds of internal supervisory review given the case's statewide significance. This is Appellee's first extension request, and Appellant X.AI had previously received a comparable 30-day extension, making the request symmetric. No merits arguments are made; the motion rests on Rule 31-2.2(b) and a supporting declaration from lead counsel.

Although the court will almost certainly grant this routine, unopposed request without meaningful scrutiny, the filing carries a signal worth tracking: California's own AG is publicly characterizing the trade secret and First Amendment questions at the heart of this case as novel and lacking established answers. The answering brief due July 15, 2026 will be the first substantive articulation of California's defense of its AI disclosure statute, and the doctrinal framework it advances—particularly on which level of First Amendment scrutiny governs compelled disclosure of AI training data—is likely to influence similar regulatory battles unfolding in other jurisdictions. If the AG's office later argues on the merits that these questions are well-settled in the government's favor, the framing here could surface as an inconsistency.

OPENING BRIEF submitted for filing by Appellant X.AI…

Issue: In *X.AI LLC v. Bonta*, X.AI argues that California's A.B. 2013 — which requires generative AI developers to publicly disclose detailed information about their training data, including dataset sources, data counts, copyright and licensing status, and processing methodology — unconstitutionally compels speech in violation of the First Amendment. The question is whether that disclosure regime goes so far beyond the narrow, deception-correcting mandates the Supreme Court has historically tolerated that it cannot survive even deferential constitutional review, or whether it is the kind of routine commercial transparency requirement that governments may freely impose. X.AI also raises a Fifth Amendment challenge, arguing that compelling disclosure of proprietary training-data details constitutes an uncompensated taking or an arbitrary deprivation of property.

X.AI LLC filed this opening brief as Appellant in the Ninth Circuit, challenging a district court ruling that left California's A.B. 2013 (Civil Code §§ 3110–3111) standing. DktEntry 15.2 is the statutory and constitutional text addendum accompanying that brief — a formal component of the opening brief package that reproduces the full text of A.B. 2013 alongside the First and Fifth Amendments. By placing the statute's detailed disclosure requirements directly against the constitutional provisions in the appellate record, X.AI invites the Ninth Circuit to assess the law's breadth without the softening effect of legislative findings or agency interpretation. The main brief argues that A.B. 2013's mandates — covering training-data sources, data volumes, licensing arrangements, synthetic data use, and consumer data presence — are far too extensive and methodologically complex to qualify as the "purely factual, uncontroversial" disclosures that receive deferential treatment under *Zauderer v. Office of Disciplinary Counsel*. X.AI seeks reversal and, presumably, an injunction against enforcement.

This case is an early test of how First Amendment compelled-speech doctrine applies to AI transparency legislation, a category of regulation that is proliferating rapidly at the state level. The central doctrinal battleground — whether training-data disclosure mandates fall within *Zauderer*'s deferential framework or demand heightened scrutiny under *NIFLA v. Becerra* — is genuinely unsettled, and a Ninth Circuit ruling will carry significant weight for how similar statutes in other states are drafted and litigated. If X.AI successfully argues that characterizing datasets, licensing arrangements, and training methodology requires expressive judgment rather than mere factual reporting, it could substantially narrow the space in which governments may regulate AI transparency without triggering serious constitutional review. Conversely, if California prevails under *Zauderer*, it would confirm broad legislative latitude to compel disclosure from AI developers, potentially accelerating similar laws nationwide.

EXCERPTS OF RECORD submitted for filing by Appellant… — Attachment 16

Issue: In *X.AI LLC v. Bonta*, X.AI argues that California AB 2013 — which requires generative AI developers to publicly post summaries of their training datasets across twelve enumerated categories — unconstitutionally compels disclosure of proprietary trade secrets and commercially sensitive technical information. The case presents the hard question of whether a state-mandated AI training-data disclosure regime triggers demanding First Amendment scrutiny as compelled speech, or whether it qualifies as ordinary commercial-speech regulation subject to the more permissive *Central Hudson* intermediate-scrutiny standard. The answer is not obvious because AI training-data summaries arguably occupy contested ground between the corrective-disclosure context in which courts are most tolerant of compelled speech and the forced revelation of proprietary editorial and technical judgments in which courts are least tolerant.

X.AI filed these Excerpts of Record in the Ninth Circuit on May 14, 2026, as the mandatory record compilation accompanying its Opening Brief on appeal under Ninth Circuit Rule 30-1. The filing assembles the trial-court record — including the March 4, 2026 district court order denying X.AI's motion for a preliminary injunction, the hearing transcript, supporting declarations, and the State's opposition — that will frame the appellate court's review. The district court, applying the *Winter* preliminary injunction standard, denied relief on all three constitutional theories X.AI advanced: a Takings Clause claim failed for lack of particularized identification of X.AI's own trade secrets; a First Amendment compelled-speech claim failed because AB 2013 likely survives *Central Hudson* intermediate scrutiny at this early stage; and a vagueness challenge failed in part because X.AI's own pleadings used the statute's contested terms fluently. The Excerpts of Record preserve those rulings and the underlying evidentiary record for de novo legal review at the Ninth Circuit.

This is the first appellate test of a state-level generative AI training-data disclosure mandate, and the Ninth Circuit's resolution of the *Zauderer*-versus-*Central Hudson* boundary in this context will carry significant weight as other jurisdictions consider similar AI transparency legislation. X.AI's most viable appellate argument centers on First Amendment proportionality: the district court itself signaled that the "limited utility of high-level dataset summaries for important consumer decisionmaking" is a genuinely open question that a fuller evidentiary record could resolve differently. If X.AI can persuade the Ninth Circuit that AI training-data disclosures are more analogous to compelled revelation of proprietary judgments than to corrective commercial disclosures — distinguishing the pharmaceutical pricing precedent the State relies on — the case could constrain how California and other states may structure AI transparency requirements going forward.