Artificial intelligence already helps legal professionals organize documents, summarize information, and conduct preliminary research. However, these advantages also create risks when software produces inaccurate citations, misleading conclusions, or exposes confidential information.
California’s new law addresses these concerns while preserving lawyers’ professional responsibilities. It does not prohibit every use of AI in legal work.
This guide explains the new requirements, their relationship to existing legal ethics rules, and what clients should understand when technology assists with their cases.
California SB 574 AI Rules: What Changes in 2027?
Senate Bill 574 establishes statutory safeguards for attorneys using generative artificial intelligence. The legislation also addresses how courts and arbitrators handle AI-assisted work.
Importantly, the law focuses on professional accountability rather than requiring lawyers to abandon useful technology.
Attorneys may continue using appropriate AI tools, but they remain responsible for their work product, legal judgment, and protection of client information.
New Responsibilities for Lawyers Using Generative AI
The legislation adds Section 6068.1 to California’s Business and Professions Code. It establishes duties concerning confidentiality, output verification, and disclosure.
These requirements apply when attorneys use generative AI to assist in practicing law. They supplement existing obligations involving professional competence and diligence.
Mandatory AI Disclosure in Court Documents
One important change involves disclosure when generative AI contributes to documents submitted to a court.
Under SB 574, an attorney using generative AI must disclose that use to the court for court-submitted documents.
For example, a lawyer using an AI system to draft part of a motion must consider the statutory disclosure requirement before filing.
However, disclosure does not mean the generated content becomes automatically reliable or admissible.
The attorney must still review the document and ensure it accurately presents the relevant facts and law.
Additionally, individual courts may establish procedural requirements concerning the format of AI disclosures.
Law firms should therefore review both the statewide statute and any applicable local court rules before submitting documents.
Disclosure also differs from client communication. The new provision separately directs attorneys to consider whether AI use should be disclosed when generating public-facing content.
Verifying Court Citations and Preventing AI Hallucinations
Generative AI can produce convincing legal language while inventing cases, misstating statutes, or creating inaccurate quotations.
These errors are commonly described as AI hallucinations.
SB 574 responds by requiring attorneys to take reasonable steps to verify generated information and correct erroneous output.
For example, a legal memorandum might include a case citation that appears authentic. Before relying on it, an attorney should locate the actual decision and confirm its meaning.
The legislation also strengthens responsibility for citations included in court filings.
Lawyers cannot simply assume that a citation is accurate because it appears in a legal research summary or automated document.
Importantly, independent verification remains necessary even when technology presents its answer with confidence.
These safeguards can benefit clients by reducing the risk of misleading arguments, unnecessary expenses, and procedural complications caused by inaccurate filings.
Client Confidentiality and Professional Judgment
Client confidentiality is fundamental to the attorney-client relationship. Legal matters often involve sensitive medical information, financial records, private communications, and personal identifying details.
Generative AI systems create additional questions about where submitted information travels, who can access it, and whether it may be retained.
Protecting Confidential Information in AI Tools
SB 574 restricts attorneys from entering protected client information into generative AI systems when access is not appropriately limited to authorized persons subject to confidentiality obligations.
This requirement is particularly important when lawyers evaluate publicly accessible technology.
For instance, entering an unredacted medical record into an unsuitable service could expose confidential information beyond the authorized legal team.
Similarly, documents containing addresses, financial information, account numbers, or sensitive communications require careful handling.
Law firms should evaluate contractual protections, access controls, retention practices, and confidentiality safeguards before using an AI platform.
Removing identifying information may reduce certain risks, although anonymization does not automatically resolve every confidentiality concern.
Ultimately, responsibility for protecting information remains with the attorney and legal organization.
Why Human Lawyers Must Retain Legal Judgment
Although generative AI can organize material and suggest language, it cannot independently assume the professional responsibilities of a licensed attorney.
SB 574 expressly prevents attorneys from delegating the practice of law to generative artificial intelligence.
This distinction matters because legal representation requires more than generating plausible text.
Attorneys must evaluate evidence, consider client objectives, identify legal risks, and apply professional judgment to individual circumstances.
For example, an AI system might summarize a personal injury settlement agreement. However, a lawyer must determine whether the proposed terms adequately protect the client’s interests.
Likewise, software might identify a procedural deadline, but the attorney remains responsible for confirming the correct deadline.
The legislation also addresses arbitrators, prohibiting them from delegating their decision-making responsibilities to generative AI.
These provisions reinforce the principle that technology may support legal work without replacing accountable human decision-makers.
How California’s New AI Law Could Affect Legal Cases
The California SB 574 AI rules will matter to attorneys, law firms, businesses, and consumers who rely on professional legal services.
However, the law does not establish that every AI-assisted legal document is defective or that technology should be excluded from litigation.
Instead, it seeks to ensure that professional responsibility continues when lawyers adopt new tools.
What Clients, Law Firms, and Courts Should Prepare For
Legal organizations can prepare before the January 2027 effective date by reviewing their AI policies, staff training, document procedures, and approved technology.
Clients can also benefit from understanding how their information is managed and who remains accountable for legal decisions.
Practical preparation for law firms
Law firms should establish written procedures describing appropriate AI use.
These procedures can identify approved platforms, restricted information, verification standards, and documentation requirements.
Training should also address how lawyers distinguish AI-generated suggestions from verified legal authority.
In addition, firms may benefit from maintaining records of significant technology-assisted research and drafting activities.
However, internal documentation practices must themselves protect confidential information and comply with professional obligations.
What clients should ask their attorneys
Clients may reasonably ask how generative AI is used in their legal matters.
Useful questions include whether confidential information enters third-party software, whether legal citations receive independent review, and who approves final court submissions.
These questions encourage transparency without assuming that AI use necessarily indicates poor representation.
In fact, properly supervised technology can help professionals manage complex information more efficiently.
Nevertheless, efficiency should never replace sound judgment, accurate research, or secure handling of sensitive records.
How SB 574 relates to AI-generated evidence
The new law concerns lawyers’ professional use of generative AI. That differs from determining whether an AI-generated photograph, recording, or document may be admitted as evidence.
Evidence disputes involve questions about authenticity, reliability, relevance, and applicable procedural rules.
For example, a manipulated video may require forensic examination even when an attorney properly discloses their own use of AI.
Likewise, an authentic recording does not become unreliable simply because another party alleges it was artificially generated.
Our related article, AI-Generated Evidence in Court: What Changes in 2026?, examines these issues in greater detail.
Together, these developments show why accurate research, digital provenance, and professional oversight are increasingly important.
Existing California court AI policies
California courts already operate under additional AI governance requirements.
Rule 10.430 of the California Rules of Court requires courts permitting specified generative AI use by judicial officers or staff to maintain appropriate policies.
These policies address confidentiality, permissible uses, and other safeguards.
The requirements are distinct from SB 574, although both reflect concerns about responsible technology use.
Readers can review California Rule of Court 10.430 for official information about court AI policies.
Frequently asked questions
When does California SB 574 take effect?
The legislation was signed September 30, 2026, and its new statutory provisions are scheduled to take effect January 1, 2027.
Does SB 574 ban lawyers from using ChatGPT or other AI tools?
No. It establishes safeguards governing confidentiality, verification, disclosure, and professional accountability when attorneys use generative AI.
Must lawyers verify citations produced by AI?
Yes. Attorneys remain responsible for verifying legal authorities and correcting inaccurate generated material. The amended filing requirements also address verification of citations more broadly.
Can an AI system make legal decisions for a client?
Attorneys cannot delegate the practice of law to AI. Professional legal judgments must remain under the responsibility of licensed lawyers.
Does AI disclosure automatically make a document admissible?
No. Disclosure and evidentiary admissibility involve different legal questions. Applicable evidence and procedural standards still govern.
Final thoughts
California SB 574 AI rules represent an important development in legal technology regulation.
Beginning in 2027, attorneys using generative AI will face specific requirements involving court disclosures, citation verification, confidentiality, and independent judgment.
For consumers, the central message is straightforward: technology may assist legal professionals, but it does not replace their obligations to clients.
As AI adoption expands, legal organizations should prioritize clear policies, secure technology, accurate research, and meaningful human oversight.
For an official overview of the legislation, read the California Governor’s September 2026 AI legislation announcement.
Disclaimer: This article provides general legal education as of October 2026. It does not constitute individualized legal advice. Readers should confirm the operative statutory language and applicable court rules before relying on procedural requirements.