On September 30, Governor Newsom signed SB 574, thus creating—for the first time—statutory guardrails on attorneys’ and arbitrators’ use of generative artificial intelligence (AI). The law, which comes into effect in January 2027, imposes new verification requirements on lawyers while requiring that certain uses of generative AI be disclosed.

The law leaves unresolved some of the thornier issues affecting lawyers’ use of AI tools—such how privilege or work-product protection extends to AI generated materials. That said, the law reflects enhanced scrutiny of the legal community uses of these tools and a potential appetite for more aggressive use of sanctions against attorneys who fail to verify citations, correct erroneous output, or protect nonpublic information.  For law firms, legal departments, and dispute-resolution providers, SB 574 therefore underscores the importance of meaningful human review and effective controls governing generative AI use. This article examines broader context around SB 574’s passage, and its principal requirements for attorneys and arbitrators.

Response to AI Hallucinations

SB 574 comes on the heels of several court decisions imposing sanctions on lawyers for failing to check citations. Most significantly, in Malkeet Lnu v. Blanche, attorneys filed Ninth Circuit briefs containing “multiple nonexistent cases, misattributed quotations, and gross representations of real cases,” then denied using AI and described the errors as typographical. 177 F.4th 1014, 1020, 1024-31 (9th Cir. 2026). The Ninth Circuit issued a disciplinary warning that lawyers must understand the risks of generative AI, read everything cited in court filings—AI-assisted or not—and promptly disclose hallucinated authorities. Id. at 1020. When imposing sanctions, the court said they were not for using AI, but for failing to comply with procedural and ethical duties that apply regardless of how legal papers are prepared. Id. at 1024, 1026.

In response to concerns about AI hallucinations, California’s State Bar has also issued explicit guidance to lawyers on the use of AI.  State Bar of Cal. Standing Comm. on Prof’l Resp. & Conduct, Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law 1-8 (2026). The guidance emphasizes that an attorney remains responsible for AI-assisted work product and must independently review, verify, and correct AI-generated outputs.

New Verifications Duties

SB 574 codifies several aspects of this prior guidance but goes further by imposing more specific verification obligations. Specifically, it creates two related but distinct verification duties.

• First, an attorney using generative AI must take reasonable steps to (a) verify the accuracy of its outputs, expressly including case and statutory citations; and (b) correct erroneous or hallucinated output in any material she uses. § 6068.1(a)(3)(B). 
• Second, and more broadly, every citation included in a paper filed in court must be personally verified by an attorney responsible for its submission—regardless of whether that citation originated with AI. Civ. Proc. Code § 128.7(b)(2)(A). 

This second requirement is not limited to documents substantially drafted by AI or containing AI-generated citations and reflects Malkeet’s emphasis on the attorney’s responsibility at the point of signing and filing.  See Malkeet Lnu v. Blanche, 177 F.4th 1014, 1025-29 (9th Cir. 2026). Courts may enforce SB 574’s citation-verification requirement through existing sanctions procedures in California Code of Civil Procedure § 128.7.  When sanctions are sought by motion, that provision generally permits counsel 21 days to withdraw or correct challenged material.  Cal. Civ. Proc. Code § 128.7(c)(1).

New Disclosure Requirements

SB 574 requires an attorney to disclose any use of generative AI to the court for documents submitted to the court.  Cal. Bus. & Prof. Code § 6068.1(a)(3)(C).   For AI-assisted content provided to the public, SB 574 imposes a less categorical obligation: an attorney must “consider whether to disclose” the use of generative AI.  Id.

SB 574 also prohibits an arbitrator from relying on AI-generated information outside the record without first making appropriate disclosures to the parties and, as far as practical, allowing them to comment on its use.  Cal. Civ. Proc. Code § 1282.1(b)(2).  The law does not specify what an “appropriate” disclosure must contain or when permitting party comment may be impractical.  Nevertheless, this requirement is broadly consistent with the AAA-ICDR Guidance on Arbitrators’ Use of AI Tools, which advises arbitrators to disclose their use of generative AI when that use materially affects the arbitral process or the reasoning underlying their decisions.  AAA-ICDR, Guidance on Arbitrators’ Use of AI Tools 1 (Mar. 2025).

No Impermissible Delegation 

SB 574 provides that lawyers may not “delegate the practice of law” to generative AI—but does not define that prohibition.  Cal. Bus. & Prof. Code § 6068.1(a)(2).  Other than the duties discussed above, SB 574 itself leaves unresolved the point at which permissible AI assistance becomes impermissible delegation. California’s 2026 Practical Guidance provides a more concrete limit.  Specifically, it states that attorneys must not deploy agentic AI systems to make substantive legal determinations, communicate legal advice, prepare and file pleadings, or otherwise act in a representative capacity without meaningful attorney supervision and review.  State Bar of Cal. Standing Comm. on Prof’l Resp. & Conduct, Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law 2 (2026).  “Critically, any use of AI must not diminish or abdicate professional judgment. A lawyer remains fully responsible for any outputs and work product generated with the assistance of AI.”  Id. at 3.  Courts have drawn a similar line.

SB 574 likewise prohibits an arbitrator from “delegating any part of their decision-making process” to a generative AI tool.  Cal. Civ. Proc. Code § 1282.1(b)(1).  The law does not define “decision-making process” or identify tasks that constitute prohibited delegation.  Outside California, guidance from the American Arbitration Association and International Centre for Dispute Resolution (AAA-ICDR) provides that arbitrators must retain complete control over decision-making and may use AI to support—but not replace—their judgment and expertise.  AAA-ICDR, Guidance on Arbitrators’ Use of AI Tools 1 (Mar. 2025).  The American Arbitration Association (AAA)’s AI Led Arbitration Rules contemplate more substantial AI involvement.  Under the AAA’s framework, parties consent to an “AI Arbitrator” making a preliminary decision and preparing a draft award—which a human arbitrator then reviews, revises as needed, finalizes, and signs.  Am. Arb. Ass’n, AI Led Arbitration Rules R-2(c), R-8(a) (effective Feb. 18, 2026).  Thus, SB 574 may impose a stricter human-decision-making requirement, but it does not specify whether meaningful human review can cure AI involvement at an earlier decisional stage.    

Guidance for Judges and ADR Oversight 

Lastly, SB 574 addresses institutions governing judges and alternative dispute resolution (ADR) providers.  It requires the California Judicial Council to consider whether to update Standard 10.80, which provides nonbinding guidance for judicial officers using generative AI—including guidance regarding confidentiality, accuracy, bias, and possible disclosure of AI use. Cal. Civ. Proc. Code § 180.

***

SB 574 may be the first law regulating lawyers’ use of AI; it likely won’t be the last. More than codifying familiar ethical duties, it puts teeth behind them—and sends lawyers a clear message: AI may assist their work, but it cannot replace their training, expertise, or judgment.