New guidance, proposed rule changes on generative AI in legal practice

Last month, the Standing Committee on Professional Responsibility and Conduct (COPRAC) of the State Bar of California issued an updated version of its Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law ("Practical Guidance"), which was first published in 2023. Also under consideration are proposed changes to the Rules of Professional Conduct. The public comment period for the changes closed on May 4, 2026.

The updated Practical Guidance replaces the 2023 version and is expressly intended to address the "unique challenges presented by the use of agentic artificial intelligence." Although the proposals consist solely of revisions to Comments rather than the Rules themselves, the proposed language moves beyond general caveats of using AI. The Comments now focus on what lawyers and law firms must do when integrating increasingly autonomous systems into their practice, aiming to provide clarity on which existing duties of competence, confidentiality, supervision and candor apply when legal work is generated, assisted or influenced by AI systems.

The Practical Guidance additionally highlights the growing trend in the legal profession in which lawyers and their clients mistakenly rely on "hallucinated" results, that "generate plausible but inaccurate information." The guidance emphasizes that lawyers remain responsible for all work product generated with the assistance of AI systems, including factual assertions, legal analysis, citations, summaries and strategic recommendations.

Expounding on this trend, the Practical Guidance distinguishes traditional AI systems and generative AI systems that create new content in response to user prompts. Such a distinction matters, as generative AI tools are increasingly utilized to draft pleadings, summarize discovery, analyze contracts, streamline legal research and generate client communications. By distinguishing between the two, COPRAC aims to frame AI as an assistive tool rather than a substitute decision maker.

This clarification was necessary, as many AI systems are no longer operating as passive drafting tools. Many large language model (LLM) platforms are built on deep neural networks, which make them capable of independently retrieving information, proposing (often hypothetical) citable authority, and generating substantial work product with minimal human involvement. Under the Rules of Competence, lawyers are required to understand "the nature and extent of an agentic system's autonomy, the circumstances under which it may act without prompting, the data sources it may access, and the safeguards necessary to prevent errors and ethical violations." Within this context, "agentic" capabilities "can enable systems to autonomously perform tasks or workflows without human prompting."

COPRAC's updated guidance makes it clear that lawyers cannot ethically delegate legal judgment to those systems simply because the technology appears to be efficient or sophisticated. Rather, lawyers must proactively monitor appropriate AI use. This includes taking "reasonable steps to understand how a generative AI product collects, uses, stores, and discloses information provided by the user." In practice, that obligation appears to require more than relying on vendor assurances or generic privacy statements. Lawyers must evaluate whether client information entered into AI systems may be retained, reused, exposed to third parties or incorporated into future model training.

Further, lawyers must retain control over strategic decision-making and exercise professional judgment when representing a client. To that end, the guidance focuses heavily on operational safeguards within law firms and legal organizations. These include establishing internal AI-use policies, limiting access to confidential information, supervising lawyers and staff using AI systems, implementing personnel protocol, and implementing review procedures designed to identify inaccurate or hallucinated output before it is incorporated into legal work product.

Of paramount importance is a lawyer's duty of candor to the tribunal. While lawyers have always been obligated to verify citations before submitting briefs, the proliferation of hallucinated authorities has heightened scrutiny from courts and opposing counsel alike.

The potential rule changes reflect the above practical guidance:

The proposed comment to Rule 1.1 (Competence) would provide: "When using technology, including artificial intelligence, a lawyer must independently review, verify, and exercise professional judgment regarding any output generated by the technology that is used in connection with representing a client." Rather than prohibiting AI use, the proposed comment clarifies that lawyers remain personally responsible for reviewing and validating AI-generated work before it is used in representation.

The proposed Comment to Rule 1.4 (Communication with Clients) would require lawyers to communicate sufficient information when the use of AI "presents a significant risk or materially affects the scope, cost, manner, or decision-making process of representation." Practically speaking, lawyers may need to evaluate whether AI use changes the nature of the representation in ways clients would reasonably want to understand, including how work is performed, whether confidential information is processed through third-party systems, or whether AI use impacts staffing, billing or litigation strategy.

The proposed Comment to Rule 1.6 (Confidential Information of a Client) would clarify that "reveal" includes exposing confidential information to AI tools where there is a material risk that the information "may be accessed, retained, or used" inconsistently with the lawyer's confidentiality obligations. This proposed clarification is particularly important given the widespread use of public-facing AI tools. The guidance makes clear that inputting internal confidential information into external learning models may itself create disclosure risks if reasonable safeguards are not in place.

The proposed Comment to Rule 3.3 (Candor Toward the Tribunal) would expressly require lawyers to verify the "accuracy and existence of cited authorities," including those "generated or assisted by artificial intelligence or other technological tools."

Also under consideration are comments to Rules 5.1 and 5.3 addressing managerial and supervisory obligations when technology is used in the provision of legal services. Those proposed comments reinforce that ethical AI use does not fall solely on an individual lawyer. Instead, the guidance suggests law firm leadership and supervising attorneys may bear responsibility for implementing reasonable oversight structures governing how AI systems are used throughout an organization.

As a caveat, while not addressed in the proposed changes, lawyers should be mindful to discuss with their clients that the client's AI-generated summaries/notetaking of meetings or discussions with the lawyers may waive the attorney-client privilege and ensure that the clients understand and agree that they should not use generative AI to alter documents, data, or evidence in the matter.

Finally, practitioners should keep an eye on SB 574, which proposes similar obligations for lawyers using generative artificial intelligence. The current version of the bill would require lawyers "to ensure that confidential personal identifying, or other nonpublic information, is not entered into a public generative artificial intelligence system" and "to ensure that reasonable steps are taken to verify the accuracy of generative artificial intelligence material and to correct any erroneous or hallucinated output in any material used by the attorney." California regulators appear to be signaling that the ethical risks associated with AI use are extensions of existing professional obligations, not entirely new categories of misconduct. As these technologies become more integrated into legal practice, the central issue will likely not be whether lawyers use AI at all, but whether they understand how to use it competently and ethically.

The Rosing Pott & Strohbehn LLP Ethics and Risk Management Team writes a monthly legal ethics column with practical insights to assist California Practitioners understand cutting edge ethics issues, manage risk, and ensure compliance. More about the Team and the authors--Heather Rosing, Dave Majchrzak, Christine Rosskopf, Joanna Mishler, and Susie Dent--can be found at https://rosinglaw.com/people/.