The legislation, now pending in the California Assembly, would set strict duties for attorneys using generative AI tools.
08/10/2026 1:00 P.M.
California lawmakers are considering legislation to establish strict operational standards for attorneys and arbitrators who use generative artificial intelligence tools.
Senate Bill 574, introduced by State Rep. Tom Umberg, D-Calif., passed the Senate in a 39-0 vote in January 2026 and is under review in the state Assembly.
Bloomberg Law reported that the measure faces an Aug. 13 deadline to clear the legislature’s fiscal appropriations committees before advancing to full floor votes in both chambers ahead of the Aug. 31 legislative session deadline. If passed by both chambers before the end of August, the bill will go to Gov. Gavin Newsom for signature or veto.
S.B 574 addresses growing concerns over AI hallucinations, confidentiality leaks, and automated decision-making in legal proceedings. For law firms and attorneys operating in California, the bill introduces specific boundaries regarding client data privacy and document verification.
Under proposed Section 6068.1 of the Business and Professions Code, attorneys using generative AI would have a statutory duty to prevent confidential, personally identifiable, or nonpublic information from being entered into public generative AI platforms. The restriction covers Social Security numbers, financial records, and private contact information.
Attorneys would also be required to take reasonable steps to verify the accuracy of AI-generated work, correct wrong or hallucinated outputs, and remove biased or harmful content — including work prepared by third parties on their behalf.
Additionally, the bill would amend Section 128.7 of the Code of Civil Procedure, requiring the attorney responsible for submitting a court filing to personally read and verify each cited source.
ACA’s Take
The intent behind S.B. 574 aligns with growing concern across the legal and accounts receivable management (ARM) industries over the risks of generic AI models.
As AI hallucinations continue to surface in legal proceedings — appearing in more than 1,600 court filings nationwide, according to The National Law Journal — unverified outputs present clear liability risks. Cases such as McCormick v. Texakoma Financial and Barnes v. Williams demonstrate how federal courts are identifying fabricated case citations generated by unverified AI tools. In Grimmer v. Citibank, the court issued sanctions after determining cited precedents and legal arguments did not exist.
Allowing your employees to use generic AI tools to research compliance questions is “a catastrophic compliance disaster waiting to happen,” ACA CEO Scott Purcell told attendees at ACA’s 2026 Annual Convention & Expo in July.
To manage compliance risks while leveraging technology, agencies should establish strict policies directing compliance staff to use vetted and trusted resources — and only those resources.
Related Content from ACA:
Don’t Get Your Compliance Info from ChatGPT
Remember, subscribe to ACA Daily and Member Alerts under your My ACA Assistant profile when logged in to acainternational.org.
The legislation, now pending in the California Assembly, would set strict duties for attorneys using generative AI tools.





