California has become the first state to regulate lawyers’ use of generative artificial intelligence by statute. Gov. Gavin Newsom signed Senate Bill 574 on September 30, 2026, and its new rules take effect January 1, 2027. The law bars attorneys from handing the practice of law over to AI tools. It also requires lawyers to verify AI output and to personally verify every citation in a court filing, to disclose AI use to the court, and to keep confidential client information out of AI systems that are not access-restricted. Arbitrators face new limits as well.
SB 574, authored by Sen. Thomas Umberg (D-Santa Ana), was one of 13 AI-related bills the Governor signed the same day. The Governor’s announcement described the measure as keeping lawyers responsible for practicing law by barring them from fully turning over core legal work, such as drafting briefs or exercising legal judgment, to AI. The bill responds to a steady stream of court decisions sanctioning lawyers for filing briefs with fabricated, AI-generated case citations.
The core duties appear in a new Business and Professions Code section 6068.1, which is part of the State Bar Act and applies to all California-licensed attorneys, including in-house counsel. The section provides that an attorney may use generative AI to assist in the practice of law but may not delegate the practice of law to it. Existing duties of competence and diligence are expressly preserved.
On confidentiality, attorneys may not enter confidential, personal identifying, or other nonpublic information into a generative AI system unless access to it is limited to the attorney and people the attorney has authorized who are themselves obligated to protect the information. The law defines personal identifying information broadly to include dates of birth, Social Security and driver’s license numbers, addresses and phone numbers of parties and witnesses, medical or psychiatric information, financial information, account numbers, and sealed or otherwise confidential material. In practical terms, commentators read this as keeping client data out of public, consumer-grade AI tools.
Attorneys using AI must also take reasonable steps to verify the accuracy of its output, including all case and statutory citations, and must correct any erroneous or “hallucinated” output in material they use. That verification duty is not limited to litigation. Finally, attorneys must disclose their use of generative AI to the court for all documents submitted to the court, and must consider whether to disclose AI use in content they provide to the public. The law defines generative AI as a system that can generate derived synthetic content that emulates its training data. That definition covers drafting and research tools but not conventional search engines, predictive analytics, or automated data processing.
Separately, the bill amends Code of Civil Procedure section 128.7, California’s counterpart to federal Rule 11. As amended, any brief, pleading, motion, or other paper filed in any California court may not contain citations that a responsible attorney has not personally verified. That requirement applies whether or not AI supplied the citation, and regardless of where the attorney is licensed. Violations are enforced through section 128.7’s existing sanctions procedures. These include a 21-day safe-harbor period to withdraw or correct a challenged paper, sanctions limited to what is needed to deter repetition, and joint responsibility of the law firm for violations by its partners, associates, and employees absent exceptional circumstances.
A new Code of Civil Procedure section 1282.1 bars arbitrators from delegating any part of their decision-making to a generative AI tool. It also bars them from relying on AI-generated information outside the record without first making appropriate disclosures to the parties and, where practical, letting the parties comment. The bill also amends Business and Professions Code section 6173, which concerns State Bar oversight of alternative dispute resolution providers. A new Code of Civil Procedure section 180 directs the Judicial Council to consider whether to update its nonbinding judicial administration standard on judges’ use of generative AI. The Assembly committee analysis of the bill describes its purpose as codifying safeguards for those receiving legal services.
Law firm commentary has broadly agreed that the law codifies duties lawyers already had, while adding statutory teeth and leaving several key terms undefined. The most frequently flagged gap is that the statute never defines what it means to “delegate” the practice of law. The law does not resolve how attorney-client privilege and work-product protection apply to AI-generated material. It further suggests that the arbitrator provisions may be stricter than the American Arbitration Association’s rules for AI-led arbitration, under which an AI drafts a preliminary award for a human arbitrator to review.
King & Spalding flags two drafting questions. First, read literally, the disclosure clause could require an attorney who used AI for anything to say so in every court filing. The firm suggests the better reading limits disclosure to AI use in preparing the particular document. Second, the statute does not say which attorney is “responsible” for verifying citations when several lawyers work on a filing. The firm adds that handing the check to a second AI tool is unlikely to satisfy the personal-verification requirement. It advises companies to make sure their outside-counsel billing guidelines do not discourage the required checks, and to configure internal AI tools to protect confidential information. The Metropolitan News-Enterprise characterized the law as codifying warnings already issued by courts and the State Bar.