SACRAMENTO, California — California is preparing to introduce a more enforceable framework for artificial intelligence in the legal profession after Governor Gavin Newsom signed Senate Bill 574 (SB 574) on September 30, 2026. The legislation establishes statutory responsibilities for attorneys and arbitrators who use generative artificial intelligence, including requirements involving confidential information, verification of AI-generated material, and disclosure to courts.
The new requirements are scheduled to take effect on January 1, 2027, making the remaining weeks of 2026 an important preparation period for California law firms, corporate legal departments, independent attorneys, and alternative dispute resolution professionals.
The legislation represents a significant development in the relationship between artificial intelligence and professional legal accountability. Rather than allowing lawyers to transfer substantive legal responsibilities to automated systems, California's new framework preserves the attorney's obligation to exercise independent professional judgment.
The development also builds on earlier measures adopted by California's judicial system and the State Bar's practical guidance on responsible AI use. However, the distinction between advisory guidance, existing court rules, and newly enacted legislation remains important.
SB 574 does not make artificial intelligence illegal in legal practice. Instead, it establishes conditions under which legal professionals may use the technology while remaining accountable for their work.
California Establishes a Statutory Framework for Legal AI
The central development is the enactment of SB 574, authored by California Senator Tom Umberg. The bill addresses attorneys, arbitrators, judicial officers, and alternative dispute resolution providers.
The California Legislature approved the measure before sending it to Governor Gavin Newsom. The governor signed it on September 30, 2026, and the Secretary of State recorded it as Chapter 858 of the Statutes of 2026.
The legislation adds Section 6068.1 to the California Business and Professions Code and amends provisions of the Code of Civil Procedure. These changes establish specific responsibilities for professionals using generative AI in legal work.
The law's scheduled effective date is January 1, 2027.
From professional guidance to enforceable obligations
Before SB 574, California lawyers were already expected to comply with professional responsibilities involving competence, confidentiality, diligence, and honesty toward courts.
The State Bar of California had also published practical guidance explaining how those responsibilities apply to generative AI. Its updated guidance was approved by the Board of Trustees on May 14, 2026.
SB 574 takes the regulatory framework further by placing several AI-related responsibilities directly into state legislation.
This distinction matters because professional guidance explains how existing obligations should be understood, while a statute establishes legal requirements through the legislative process.
The legislation therefore represents a statutory development rather than simply an administrative revision of the State Bar's guidance.
What SB 574 Requires From California Attorneys
One of the most important provisions of SB 574 concerns the limits of AI delegation.
Under the new Business and Professions Code Section 6068.1, an attorney must not delegate the practice of law to generative artificial intelligence.
This provision preserves the lawyer's responsibility for professional decisions even when AI systems can perform complex tasks.
Human judgment remains essential
Generative AI can assist with legal research, document preparation, information organization, and the analysis of large volumes of material.
However, the law establishes that the technology cannot replace the attorney's professional responsibility.
The practical distinction is between using AI as an assistant and allowing an automated system to assume legal responsibilities that require independent professional judgment.
For example, an attorney may use an AI tool to organize documents or prepare an initial draft. The attorney must still evaluate the legal reasoning, confirm the relevant authorities, and determine whether the resulting work is appropriate for the client's circumstances.
The legislation does not provide an exhaustive list of permitted and prohibited AI-assisted tasks. Consequently, the precise boundary between assistance and impermissible delegation may require further interpretation.
This issue is particularly relevant for firms developing highly automated legal services.
Confidentiality Rules Place Greater Responsibilities on Law Firms
Another central element of SB 574 concerns the protection of confidential, personally identifying, and other nonpublic information.
Generative AI systems can process substantial amounts of information, but their data-handling arrangements differ considerably.
Some systems may retain user inputs, use submitted information for additional processing, or provide access through arrangements involving third parties.
These differences create potential confidentiality risks when attorneys submit sensitive client information.
Restrictions on entering protected information
Under Section 6068.1, attorneys cannot enter protected information into a generative AI system when access to that information is not restricted to the attorney and authorized individuals who are subject to confidentiality obligations.
The legislation identifies several categories of protected information, including certain personal identifying, financial, medical, and other nonpublic records.
The provision makes the security arrangements of AI systems directly relevant to professional compliance.
Law firms will therefore need to evaluate their AI providers, contractual arrangements, access controls, and information-handling procedures.
A tool's ability to generate accurate answers is not sufficient by itself to establish that it is appropriate for confidential legal work.
Practical implications for legal technology providers
Technology providers serving attorneys may face increased questions about data retention, access restrictions, security controls, and contractual confidentiality protections.
Law firms should distinguish between public consumer-facing tools and systems configured for restricted professional use.
The legislation does not prohibit every public AI product from being used for every legal task. Its restrictions depend on the type of information entered and whether the system satisfies the statutory access conditions.
That distinction is important for smaller practices that rely on commercially available AI services.
AI-Generated Legal Information Must Be Verified
The accuracy of AI-generated legal information is another major focus of the new legislation.
Generative AI can produce convincing explanations and apparently authoritative references. Nevertheless, an output may contain inaccurate legal reasoning, incomplete information, fabricated citations, or references that do not support the stated argument.
These problems are particularly serious when documents are submitted to a court.
Verification and correction requirements
SB 574 requires attorneys using generative AI in legal practice to take reasonable steps to verify the accuracy of its outputs.
The requirement specifically includes checking case citations and statutory citations.
Attorneys must also correct erroneous or hallucinated material in work products they use.
This responsibility extends beyond checking whether a citation exists. Legal professionals must consider whether the authority is accurately represented and relevant to the legal proposition being advanced.
The legislation also amends Code of Civil Procedure Section 128.7 to require the attorney responsible for submitting a court filing to personally verify its citations.
Importantly, the personal verification requirement is not limited to citations generated by artificial intelligence. It applies to citations regardless of their source.
Consequently, legal research teams may continue to use associates, paralegals, research platforms, and AI-assisted systems, but the responsible submitting attorney must undertake the required personal verification.
This creates a clearer accountability structure for legal documents.
Court Disclosure Requirements Introduce New Transparency Standards
SB 574 also addresses transparency when attorneys use generative AI in documents submitted to courts.
The statute requires disclosure of generative AI use to the court for documents submitted to the court.
This requirement is significant because it introduces a specific statutory disclosure obligation into the legal profession's use of AI.
Questions about the scope of disclosure
Although the legislation establishes the disclosure requirement, important implementation questions remain.
For example, the statutory language does not fully specify the precise form of disclosure or clearly explain every circumstance in which AI involvement in a document triggers the obligation.
Legal practitioners may therefore need additional clarification concerning whether disclosure is required for every level of AI assistance or when AI contributes more directly to substantive content.
A legal analysis published by the California Lawyers Association on October 27, 2026, highlights the uncertainty surrounding the interpretation of some provisions.
The article also explains that the new requirements will require attorneys and law firms to assess their AI systems and internal compliance arrangements.
Until courts or relevant authorities provide further clarification, a cautious approach would be to establish document-specific disclosure procedures and maintain records of AI involvement.
The purpose of transparency is to preserve confidence in court submissions without treating the mere use of technology as evidence of professional misconduct.
California's Existing Judicial AI Rules Provide Important Context
SB 574 did not emerge without an existing regulatory foundation.
California had already introduced requirements governing the use of generative AI within its judicial system.
California Rules of Court, Rule 10.430, was adopted effective September 1, 2025.
It requires courts that permit generative AI use in court-related work to establish policies addressing confidentiality, accuracy, discrimination, and responsible use.
The rule prohibits court personnel from entering confidential, personally identifying, or other nonpublic information into public generative AI systems.
It also requires reasonable steps to verify AI-generated material and correct inaccurate or hallucinated content.
Judicial officers and court employees
California Standard of Judicial Administration 10.80 provides additional guidance for judicial officers using generative AI in their adjudicative roles.
The standard addresses accuracy, harmful or biased material, and the need for judicial responsibility when using AI-generated content.
These provisions are important because they demonstrate that the state had already begun establishing expectations for responsible AI use within the courts.
However, Rule 10.430 and Standard 10.80 should not be confused with SB 574.
The judicial provisions address the responsibilities of court institutions, employees, and judicial officers within their respective scopes. SB 574 establishes additional statutory obligations for attorneys and arbitrators.
Together, these measures form a broader approach to AI accountability across California's legal system.
Legal Experts Highlight Both Benefits and Regulatory Challenges
The introduction of SB 574 has generated discussion about whether a separate statute is necessary when attorneys already have professional obligations concerning accuracy and ethical conduct.
Supporters argue that explicit statutory responsibilities provide courts with additional tools to address unreliable AI-generated legal material.
Other legal professionals question whether several provisions substantially repeat responsibilities that lawyers were already required to follow.
Senator Tom Umberg's position
Senator Tom Umberg, the legislation's author and chair of the California Senate Judiciary Committee, discussed the problem of inaccurate AI-generated material in an interview with Reuters published on September 1, 2026.
He warned that lawyers continued submitting documents containing hallucinations or other anomalies and attributing those problems to AI.
Umberg emphasized that the integrity of court proceedings depends on the reliability of information submitted by litigants and their legal representatives.
In the same interview, he said that if that integrity could not be maintained, the judicial system itself would be undermined.
His comments reflect the legislation's central objective: technology may assist legal professionals, but it cannot become an excuse for inaccurate submissions.
Wayne Stacy's assessment
A different perspective was presented by Wayne Stacy, executive director of the Berkeley Center for Law and Technology at the University of California, Berkeley School of Law.
In the same Reuters report, Stacy characterized the measure as largely duplicative of existing ethical requirements governing lawyers.
His assessment raises an important legal-policy question.
If attorneys already have duties to verify legal authorities, protect confidential information, and exercise professional judgment, how much additional protection does a new statute provide?
The opposing considerations are not necessarily incompatible.
Existing professional duties may already establish a foundation for responsible conduct, while statutory provisions can make specific AI-related responsibilities more explicit and provide additional grounds for enforcement.
The extent to which the new legislation changes everyday legal practice will depend partly on how courts interpret and apply its requirements.
What Happens Before the January 2027 Effective Date
The period between enactment and implementation provides an opportunity for legal organizations to review their technology arrangements.
Although SB 574 has been signed into law, its principal new statutory obligations do not take effect until January 1, 2027.
Law firms should therefore avoid confusing enactment with the beginning of the statute's operative requirements.
Compliance priorities for attorneys and legal departments
Several practical steps are relevant during the remaining weeks of 2026.
First, firms should inventory the AI tools used by attorneys, employees, and contractors.
Second, they should review whether those tools meet the statutory requirements for protecting confidential and nonpublic information.
Third, firms should establish procedures for checking AI-generated legal analysis, factual statements, and citations.
Fourth, attorneys responsible for court filings should implement personal citation-verification procedures.
Fifth, organizations should develop a consistent method for documenting and disclosing AI use in court submissions.
Training is also important, particularly where AI systems can perform multiple tasks with limited human intervention.
The State Bar's May 2026 practical guidance specifically addresses agentic AI, which can independently initiate tasks, select tools, and execute workflows.
The guidance emphasizes that increasing automation does not remove the attorney's responsibility for professional judgment and supervision.
These preparations may be particularly important for firms using AI throughout their research, drafting, document review, and client-intake processes.
California's approach to legal artificial intelligence is moving from professional guidance toward more explicit statutory accountability.
SB 574 establishes requirements concerning professional judgment, confidentiality, verification, and court disclosure, while existing judicial rules provide a separate foundation for responsible AI use within the court system.
The legislation does not prohibit lawyers from using generative AI. Instead, it places clear responsibility on human legal professionals to supervise the technology and remain accountable for the work produced with its assistance.
As January 1, 2027, approaches, the most important question is no longer whether AI will become part of legal practice. It is whether attorneys, law firms, technology providers, and courts can establish reliable procedures that protect clients and preserve the integrity of judicial proceedings.
The effectiveness of California's new framework will ultimately depend on implementation, judicial interpretation, and the ability of legal professionals to maintain meaningful human oversight as AI systems become increasingly sophisticated.