In brief
California Governor Gavin Newsom signed SB 574 on September 30, 2026, making California the first state to regulate attorneys’ use of generative artificial intelligence (AI). At a high level, SB 574 allows an attorney to use AI to “assist in the practice of law,” but prohibits the “delegat[ion of] the practice of law” to AI.
Effective January 1, 2027, SB 574 requires attorneys to personally verify citations in court filings (regardless of AI use), review and correct AI-generated work, disclose AI use in court submissions, and safeguard confidential and personally identifying information entered into AI systems. It also bars arbitrators from delegating decision-making to AI and restricts their reliance on AI-generated information outside the record.
In the wake of SB 574, corporate legal departments should work with outside counsel to determine who will verify citations, which AI tools may be used, and how attorneys will track and disclose AI use in court filings.
Key takeaways
- Attorneys may not delegate the practice of law to AI. SB 574 prohibits attorneys from “delegat[ing] the practice of law to generative AI,” but expressly allows the use of AI to “assist in the practice of law.” However, the type of conduct that constitutes “delegation” is not defined by the bill.
- Citation verification applies even without AI. Attorneys responsible for submitting court filings must personally verify citations from any source. Research and drafting may still be delegated to AI, but the law imposes a separate verification requirement on attorneys before filing.
- AI restrictions may complicate internal investigations. The law also prohibits attorneys from entering a witness’s date of birth, address or phone number into an AI system unless access is restricted to the attorney and authorized persons bound by confidentiality. This requirement applies even where the information is public. If counsel hires any investigators, counsel must ensure they also comply with SB 574.
- Attorneys must check AI-generated work and disclose its use. California attorneys must take reasonable steps to verify AI-generated work, correct errors, and disclose AI use to courts for submitted documents. However, the law leaves questions about the form and scope of disclosure unresolved.
In more detail
Court citation verification extends beyond AI-generated output
California Code of Civil Procedure section 128.7 already requires attorneys to make a reasonable inquiry into the factual and legal support for their court filings. SB 574 imposes an additional obligation: briefs, motions, pleadings, or other filings cannot contain citations that an attorney has not personally verified. Notably, the rule applies whether the citations came from an AI tool, a legal research database, or another attorney.
This requirement applies to the attorney responsible for submitting the brief, motion, or other filing. However, the statute does not specify whether that attorney must be the signatory or how responsibility should be allocated when several attorneys participate in submitting the filing.
The law also leaves the method of verification open. An earlier draft of SB 574 required citations to be “personally read and verified,” but the Legislature removed “read” before enactment. That change eliminates an express reading requirement without removing the obligation to verify. The statute’s reference to “any citations” could also be interpreted to reach deposition testimony, exhibits, and other record evidence.
Litigation teams should identify the attorney responsible for verification before filing and determine how both legal authorities and record citations will be checked. Violations are subject to section 128.7’s existing sanctions procedures. Additionally, section 128.7’s discovery exception remains in place.
AI-generated content has a separate verification requirement
California attorneys are already subject to duties of competence, confidentiality, and professional judgment. The California State Bar addressed those obligations in relation to AI use in guidance first issued in 2023 and updated in May 2026. SB 574 adds express statutory requirements.
SB 574 creates the new Business and Professions Code section 6068.1, which prohibits attorneys from delegating the practice of law to generative AI. Accordingly, attorneys using AI must take reasonable steps to verify AI outputs, including legal citations, and correct inaccurate or hallucinated material it produces.
This standard differs from the personal citation verification rule. Section 6068.1 requires reasonable steps to check AI-generated work, whether or not it is submitted to a court. Section 128.7 requires personal verification of citations in covered court filings, even when AI played no role. Both requirements may apply when an attorney uses AI to prepare a brief.
These duties extend to any California-licensed attorney, including in-house counsel. Though the statute does not address out-of-state attorneys appearing pro hac vice in California courts, the broad scope of SB 574 can reasonably be interpreted to apply in these circumstances. Whether using AI to research a legal question or draft a contract, attorneys must evaluate the resulting work rather than rely on the tool’s conclusions without appropriate review.
Publicly available witness information may still be restricted
Section 6068.1 also prohibits attorneys from entering confidential, personally identifying, or other nonpublic information into generative AI systems unless access to that information is restricted to the attorney and authorized persons subject to confidentiality obligations.
The scope expressly includes information that attorneys routinely use in internal investigations: dates of birth; addresses and telephone numbers of witnesses and parties; and medical and financial information. Notably, there is no general exception for personally identifying information obtained from public records.
This could affect civil litigation and internal investigations. Attorneys often use a witness’s date of birth or address to distinguish people with similar names, locate prior proceedings, and investigate credibility. Entering that information into an AI system without the required access and confidentiality protections could violate the statute, even when the information is publicly available.
SB 574 does not prohibit ordinary public records research or all AI-assisted investigations. It does, however, prohibit attorneys from entering protected information into AI systems unless access is limited to the attorney and authorized persons bound by confidentiality obligations. The statute does not define what constitutes restricted access, and how that requirement will be interpreted for different AI systems remains to be determined.
Retaining an outside investigator does not absolve attorneys of this requirement. Section 6068.1 directly regulates attorneys, while California Rule of Professional Conduct 5.3 separately requires attorneys to supervise nonlawyer assistants, including investigators retained as independent contractors. An attorney may be responsible for conduct they direct or knowingly approve, in addition to failing to address certain conduct when the rule requires action. However, the statute does not automatically make counsel responsible for every AI search an investigator independently performs.
Legal departments and litigation teams should therefore establish protocols regarding the use of AI and what information may be entered into each tool. They should also give outside investigators clear instructions concerning protected witness information. An enterprise AI subscription alone does not establish compliance — attorneys have an obligation to confirm who can access the information submitted and what confidentiality obligations apply.
Court disclosure requirements leave practical questions open
Section 6068.1 requires attorneys using generative AI to disclose that use to courts for all documents submitted to the court. It does not specify how disclosure must be made or clearly identify which uses require it.
Using AI to draft a legal argument presents a different question from using it to organize research, summarize documents, or edit language, for example. And the statute does not explain whether each activity triggers the same disclosure obligation.
This uncertainty may arise when several attorneys contribute to a filing. For example, an associate may use AI for research or drafting without the attorney responsible for submission knowing how the technology was used. As such, litigation teams should establish protocols to assist with determining the appropriate disclosure under the statute and any applicable court orders. Disclosures should also avoid revealing privileged communications or litigation strategy.
By contrast, the statute treats public-facing materials differently. Attorneys who use AI to create content provided to the public must consider whether to disclose its use, but disclosure is not automatically required.
Conclusion
SB 574 takes effect January 1, 2027. Before then, businesses should coordinate citation verification and court disclosure procedures with outside counsel and determine which AI tools may be used for particular tasks. Baker McKenzie can assist companies in evaluating these requirements and developing procedures that account for their litigation practices and existing AI tools.