Guardrails for legal AI: What California's SB 574 would require of attorneys and arbitrators

SB 574 is a California bill that would set specific duties for attorneys who use generative artificial intelligence (AI) and would restrict how arbitrators may use such tools in decision-making. It would amend provisions in the Business and Professions Code and the Code of Civil Procedure to address confidentiality, accuracy, bias and citation verification for attorneys, and to prohibit delegation of arbitral decision-making to AI while adding disclosure and responsibility requirements for arbitrators.

Background

This bill was introduced by California State Sen. Tom Umberg, who represents Senate District 34, which includes northern Orange County and eastern portions of Los Angeles County. Sen. Umberg, a former practicing attorney and retired U.S. Army Colonel, stated that the bill is intended to address the growing use of AI in the legal system by providing "clear guardrails to protect clients' confidentiality and ensure that real people, not algorithms, are making legal decisions."

Although the bill partially responds to nationwide concerns regarding attorneys submitting court filings containing fictitious case citations generated by AI tools, it also reflects concerns that arbitrators could input evidence into an AI platform and rely on AI to generate a decision. 

The bill draws in part from the California Rules of Court, Standard 10.80, as well as the reasoning in recent case law related to AI.

What SB 574 would require of attorneys

The proposed bill would require attorneys to protect client confidentiality, verify all AI-generated content, and avoid discriminatory or biased outputs.

Confidentiality: Attorneys using generative AI must ensure that confidential, personal identifying, or other nonpublic information is not entered into a public generative AI system. Personal identifying information includes items such as driver's license numbers, dates of birth, Social Security numbers, National Crime Information and Criminal Identification and Information numbers, contact information of parties and court personnel, medical or psychiatric information, financial information, account numbers, and any content sealed by court order or deemed confidential by rule or statute.

Accuracy and accountability: Attorneys must take reasonable steps to verify the accuracy of AI-generated material, correct erroneous or hallucinated output, and remove biased, offensive or harmful content in any AI material they use, including work prepared on their behalf by others.

Anti-discrimination: Attorneys must ensure their use of AI does not unlawfully discriminate against or disparately impact protected classes, including categories such as age, ancestry, color, ethnicity, gender, gender expression, gender identity, genetic information, marital status, medical condition, military or veteran status, national origin, physical or mental disability, political affiliation, race, religion, sex, sexual orientation, socioeconomic status and other classifications protected by federal or state law.

Public-facing content: Attorneys must consider whether to disclose AI use when AI is used to create content provided to the public.

Citation verification in court filings: Attorneys must read and verify every citation in court filings, including citations provided by generative AI.

What SB 574 would require of arbitrators

Under the proposed bill, arbitrators cannot delegate decision-making to AI, must independently analyze evidence and must disclose any use of AI-generated information outside the record.

No delegation to AI: Arbitrators may not delegate any part of their decision-making process to generative AI, and AI use cannot replace their independent analysis of facts, law and evidence. Arbitrators must avoid delegating tasks to AI if such use could influence procedural or substantive decisions.

Limits on extra-record AI information: Arbitrators may not rely on AI-generated information outside the record without first making appropriate disclosures to the parties and, as practical, allowing comment. If an AI tool cannot cite independently verifiable sources, the arbitrator may not assume such sources exist or are described accurately.

Responsibility for awards: Arbitrators must assume responsibility for all aspects of an award regardless of any AI assistance used in the process.

SB 574's timing

SB 574 passed the California Senate on Jan. 29, 2026, by a unanimous 39-0 vote, and is currently pending in the Assembly. While a unanimous Senate approval reflects strong support, it does not guarantee passage in the Assembly. There, the bill will proceed through committee review, hearings and ultimately a floor vote. The Assembly must pass the bill by Aug. 31, 2026, the final day of the legislative session.

If the bill passes, it will be presented to the governor, who may (1) sign the bill into law, (2) allow it to become law without a signature or (3) veto it. In most cases, the governor has 12 days to act after receiving a bill. However, if the bill passes on the final day of the legislative session, Aug. 31, 2026, the governor has until Sept. 30, 2026 to sign or veto the bill; otherwise, the bill becomes law without a signature.

If enacted, the law would typically take effect on Jan. 1 of the following year.

Bottom line

SB 574 would not ban legal professionals from using AI. Rather, it would set guardrails that reinforce confidentiality, accuracy, fairness, transparency and personal accountability in both litigation and arbitration contexts.

As the bill moves through the Assembly, attorneys and arbitrators should begin preparing for a future where technology and human judgment work hand in hand.