Artificial intelligence can help lawyers draft documents, summarize information, and conduct legal research. But it sometimes produces nonexistent case citations, inaccurate statements of law, and errors that are difficult to recognize without careful review. AI can also create another problem: information supplied to an AI system may include confidential client details that should never be disclosed to unauthorized persons.
To address these concerns, California Senate Bill 574, signed into law on September 30, 2026, and effective as of January 1, 2027, establishes new statutory requirements governing the use of generative artificial intelligence by attorneys and arbitrators.
The new law also strengthens procedures for addressing ethical complaints involving certain alternative dispute resolution providers and directs the Judicial Council to revisit a standard that governs judicial administration.
For attorneys, the legislation makes clear that using AI does not transfer professional responsibility to a machine. For clients and other participants in legal proceedings, it establishes additional safeguards concerning confidentiality, accuracy, and decision-making.
One of SB 574’s central provisions prohibits an attorney from delegating the practice of law to generative AI. The legislation doesn’t prevent lawyers from using AI to assist with their work, but it draws a distinction between using technology as a tool and surrendering professional responsibilities to it.
An attorney who uses generative AI must take reasonable steps to verify the accuracy of its output, including case citations and statutory references. The attorney must also correct erroneous or “hallucinated” output in any material the attorney uses. (A “hallucination” occurs when an AI system produces information that appears credible but is false, unsupported, or fabricated.)
This matters because AI-generated legal writing can sound authoritative even when its underlying authorities do not exist or do not support the proposition for which they are cited. Under the new law, an attorney cannot excuse an inaccurate filing by explaining that an AI system generated the mistake.
SB 574 also requires an attorney to disclose to the court if generative AI has been used in documents submitted to the court.
Separately, the statute directs attorneys to consider whether to disclose the use of AI technology to create content provided to the public.
These requirements place disclosure and verification within the attorney’s professional obligations. The legislation does not prescribe a particular disclosure format or establish a method for documenting verification.
Confidentiality Is a Separate Concern
The use of AI tools can raise confidentiality issues even when their output is accurate. A lawyer might enter a client’s financial records, medical information, litigation strategy, or other sensitive material into a system to obtain a summary or draft. Depending on the system’s access controls, doing so could expose information to people who are not authorized to receive it.
SB 574 addresses this risk directly. An attorney may not enter confidential, personally identifying, or other nonpublic information into a generative AI system unless access to that information is restricted to the attorney and to persons the attorney authorizes who are subject to obligations to protect its confidentiality.
Personally identifying information identified by the law includes driver’s license numbers, dates of birth, Social Security numbers, addresses and telephone numbers of parties and witnesses, medical or psychiatric information, financial information, account numbers, and information sealed by court order or made confidential by statute or court rule.
For law firms, the practical implications extend beyond an individual lawyer’s choice of AI tool. Firms should evaluate the access controls and confidentiality protections of the systems they permit lawyers and staff to use. They should also consider how their policies address client information, review of AI-generated work, and required disclosures to courts.
For clients, the legislation reinforces an important principle: efficiency cannot come at the expense of confidentiality or the lawyer’s obligation to exercise professional judgment.
AI-Generated Citations Must Be Personally Verified
SB 574 amends Code of Civil Procedure section 128.7, which governs representations made when papers are submitted to a court.
The amended statute prohibits a brief, pleading, motion, or other paper filed in court from containing a citation that the attorney responsible for submitting it has not personally verified. The prohibition specifically includes citations supplied by generative AI.
Section 128.7 already provides a framework for sanctions when its requirements are violated, including notice and an opportunity to respond. SB 574 adds the express citation-verification requirement to that framework.
The amended statute also limits monetary sanctions against a represented party for a violation of the new citation provision.
Requirements for Arbitrators
The legislation prohibits an arbitrator from delegating any part of the decision-making process to a generative AI tool.
An arbitrator also may not rely on information generated by AI outside the evidentiary record without first making appropriate disclosures to the parties and, to the extent practical, allowing them to comment on its use.
The distinction between lawyers and arbitrators is important. SB 574 permits attorneys to use AI as an aid subject to safeguards, but it expressly bars arbitrators from delegating any part of their decision-making to it.
SB 574 also revises the State Bar’s voluntary certification program for alternative dispute resolution firms, providers, and practitioners. The program must include procedures for the State Bar, or an approved professional organization, to receive, investigate, and resolve complaints alleging violations of applicable ethical standards.
Complaint proceedings generally must remain private and confidential. Following a decision, however, the State Bar may publicly disclose certain information, including the identity of a certified provider found to have violated an ethical standard, remedial action taken, and the general basis for that action, provided confidential or privileged communications are not revealed.
The statute includes specific protections for privileged and confidential materials and limits the circumstances in which those materials must be produced. These provisions are particularly detailed for a mediation that involves a self-represented party.
Separately, SB 574 directs the Judicial Council to revisit and revise as necessary Standard 10.80 of the California Standards of Judicial Administration to reflect further developments in generative AI.
The Practical Takeaway
SB 574 does not eliminate AI from legal practice; it establishes boundaries around its use. Attorneys must protect confidential information, verify AI-generated work, correct material errors, disclose to courts the use of AI for submitted documents, and personally verify citations.
Arbitrators may not delegate decision-making to AI and must disclose qualifying reliance on AI-generated information that is outside the record.
For law firms, the prudent response is to review AI policies, approved tools, confidentiality controls, citation-checking procedures, and disclosure practices.
For clients, the legislation offers additional safeguards while underscoring the continuing importance of professional accountability.