California’s New AI Law: Implications for Lawyers and eDiscovery Teams

On September 30, 2026, Governor Gavin Newsom signed Senate Bill 574, California’s new law governing attorney use of generative artificial intelligence (AI). Starting January 1, 2027, attorneys who use generative AI will face new statutory duties under California law, and every citation in a California court filing will need to be personally verified by an attorney responsible for the filing. Much of the law codifies duties attorneys already have under the Rules of Professional Conduct and existing State Bar guidance. The most significant changes are the personal-verification requirement for citations and a mandatory duty to disclose AI use to the court. For eDiscovery teams, there is an additional issue that has received less attention: how the law applies when AI-assisted review and summaries feed into court filings.

The law has two main parts for attorneys: a set of AI duties for attorneys added to the State Bar Act (Business and Professions Code Section 6068.1) and a new citation rule added to California’s sanctions statute (Code of Civil Procedure Section 128.7). The citation rule also carries consequences for firms: absent exceptional circumstances, longstanding language in the sanctions statute makes law firms jointly responsible for violations committed by their partners, associates, and employees. The AI duties have no equivalent provision.

New Duties for Attorneys Using Generative AI

The AI duties apply whenever an attorney uses generative AI to assist in the practice of law. Three duties apply to everyday use of AI tools; the fourth duty, court disclosure, is discussed separately below.

• Professional judgment. The law contemplates that attorneys will use generative AI to assist their work but forbids “delegat[ing] the practice of law” to it and preserves existing duties of reasonable competence and diligence. AI can assist with research, drafting, and document review, but the legal judgment remains the lawyer’s, which is consistent with existing professional obligations.
• Confidentiality. Attorneys may not input confidential, personally identifying, or other nonpublic information into a generative AI system unless access to that information is restricted to the attorney and authorized persons subject to confidentiality obligations. This largely tracks attorneys’ existing confidentiality obligations and the State Bar’s 2023 guidance on generative AI. In practice, this may preclude using many consumer-grade AI tools with client information.
• Accuracy. Attorneys must take reasonable steps to verify AI outputs and correct errors in those outputs. This duty appears broader than the citation rule discussed below because it applies to AI output generally, not only to court filings.

Personal Citation Verification and Firm Responsibility

Under the new citation rule, a brief, pleading, motion, or other paper filed in a California court may not contain “any citations that an attorney responsible for submitting the pleading has not personally verified, including any citation provided by generative artificial intelligence.” Because the rule covers “any citations” and treats AI-generated citations as one category within them, it appears to reach all citations, regardless of source. The rule draws on the Court of Appeal’s 2025 statement that no filing should contain citations, whether provided by generative AI or any other source, that “the attorney responsible for submitting the pleading has not personally read and verified.” Noland v. Land of the Free, L.P., No. B331918, slip op. at 2-3 (Cal. Ct. App. Sept. 12, 2025). Fabricated citations were already sanctionable before SB 574; the Noland court imposed $10,000 in sanctions on counsel. What is new is that personal verification is now a statutory requirement rather than judicial guidance.

Courts have not yet set standards for what it means to “personally verify” the citations. Until they do, a best practice is to decide at the start of work on each filing which responsible attorney will personally verify the citations and to build time for that into the schedule.

Disclosure of AI Use

The law also requires attorneys to disclose generative AI use to the court for documents submitted to it, but it does not specify the form of disclosure or what degree of AI involvement triggers it. Before SB 574, any such obligation generally came from individual courts’ local rules or judges’ standing orders. Until courts provide guidance, a best practice is to adopt a standard disclosure statement that can be adapted for each filing. Attorneys must also “consider whether to disclose the use of generative artificial intelligence if it is used to create content provided to the public.” The phrase “content provided to the public” appears broad enough to reach law firm blogs, websites, and other public-facing materials, and possibly even materials presented to jurors, although the statute does not address any specific application.

What This Means for eDiscovery

For eDiscovery practitioners, a particularly significant application may arise when discovery-derived materials become the basis for covered court filings. AI-assisted review tools increasingly identify documents, summarize testimony, and describe evidentiary takeaways.

Citations to the factual record. The personal-verification-of-citations rule does not expressly limit the personal-verification requirement to legal authorities, so it could extend to citations to the factual record, such as exhibits and deposition testimony. That reading is not settled, however. The AI duties refer specifically to “case and statutory citations,” and a court could read the two provisions together to limit “citations” to legal authority. Either way, the duty to verify AI outputs appears to apply, and a best practice is to check any record citation drawn from an AI summary against the underlying document or transcript. And, as emphasized in Noland, verifying citations has long been a responsibility owed to the court.

Disclosing AI-assisted review. The statute does not say whether AI used upstream, in review and summarization rather than drafting, triggers the court-disclosure requirement for the resulting brief. Until courts address the issue, it is prudent to take a consistent approach to whether and how disclosures describe AI-assisted review.

Discovery responses and other discovery materials. Much like its federal equivalent, Rule 11(d), the sanctions statute excludes discovery requests, responses, objections, and motions. That exclusion predates SB 574 and remains in place, so discovery misconduct continues to fall within California’s existing discovery-sanctions framework. The AI duties, however, contain no discovery exclusion. As a result, discovery responses drafted with AI, or a privilege log prepared with AI assistance, may fall outside the citation rule but still appear to be subject to the duty to verify and correct AI outputs. Firms should therefore consider procedures that address AI use in discovery as well as in court filings.

Loading data into AI review tools. Before a client’s collection or an opposing party’s production is loaded into a generative AI review or summarization tool, firms should confirm that access to the data is restricted as the statute requires by reviewing the vendor’s access controls and data-use terms.

Will Other States Follow?

California’s legislature plays an unusually substantial role in State Bar oversight, while the California Supreme Court retains ultimate authority over admission and discipline. Because of that unusual legislative role, other states may not quickly follow suit.

New York has a pending proposal that would require a separate affidavit disclosing generative AI assistance and certifying that a human reviewed the source material and verified the accuracy of the generated content. See S. 2698, 2025-2026 Reg. Sess. (N.Y. 2025). Unlike California’s law, it does not expressly require that an attorney responsible for the filing perform the verification.

Preparing for January 1

Before the law takes effect, law firms and legal departments may wish to consider the following best practices:

• Adopting a standard court-disclosure statement that teams can adapt to each filing
• Confirming which AI tools, including eDiscovery review and summarization platforms, meet the confidentiality standard, and communicating which tools are approved for client information
• Updating eDiscovery review protocols so that any AI-generated document identification or summary relied on in a filing is checked against the underlying documents
• Extending guidance to attorneys practicing in California through pro hac vice admission, rather than assuming these issues concern only California-barred attorneys

The law sets a minimum standard. Attorneys should also check local rules and judges’ standing orders for additional AI-related requirements.

© Arnold & Porter Kaye Scholer LLP 2026 All Rights Reserved. This Blog post is intended to be a general summary of the law and does not constitute legal advice. You should consult with counsel to determine applicable legal requirements in a specific fact situation.