California's New AI Ethics Rules: What Six Rule Changes Mean for Your Practice
California just made a move that should have every attorney in the state — and every attorney watching where bar regulation is heading — paying close attention.
In March 2026, the State Bar’s Standing Committee on Professional Responsibility and Conduct (COPRAC) approved proposed amendments to six Rules of Professional Conduct, weaving AI-specific obligations into the enforceable rules for the first time. The public comment period closed May 4. The rulemaking process continues, but the direction is clear: in California, AI ethics is no longer guidance. It’s the rules.
This matters beyond California. Most states have addressed AI through ethics opinions — persuasive, but without disciplinary force. If California finalizes these amendments, it becomes the first major jurisdiction to make AI obligations formally enforceable. Other states will follow.
Here is what each proposed amendment requires, and what it means for your practice today.
Why This Is Happening Now
The rulemaking traces directly to the California Supreme Court itself. In an August 2025 letter, the court directed COPRAC to consider incorporating the bar’s 2023 Practical Guidance on generative AI into the formal rules — and specifically asked the committee to address agentic AI systems that can plan and execute tasks with minimal human oversight.
COPRAC’s approach was deliberately integrative rather than standalone: rather than drafting a single AI rule, the committee amended six existing rules to make explicit that AI sharpens duties lawyers already owe, rather than creating entirely new categories of obligation. The practical effect, however, is significant — these would be enforceable rules, not a living document with no disciplinary teeth.
Rule 1.1 — Competence: The Verification Mandate
The most consequential proposed change lives in a new comment to Rule 1.1. When using any technology, including AI, 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.”
Read that carefully. It does not say attorneys should be generally careful with AI output. It says they must independently review and verify any output used in client work. There is no carve-out for routine tasks, no exception for low-stakes matters.
This is not an aspirational standard. If finalized, it is the bar’s definition of competent practice. An attorney who submits AI-generated research, drafts, or citations without independent verification is not merely being incautious — they are potentially violating the duty of competence.
The rule also extends the existing technology-competence language to make explicit that staying abreast of “the benefits and risks associated with relevant technology” now expressly includes artificial intelligence.
Our database of more than 1,300 documented AI sanctions cases shows exactly what happens when this verification step fails: fabricated citations, misrepresented holdings, sanctions ranging from public reprimand to five-figure fines.
Rule 1.4 — Communication: Conditional Disclosure
A new Comment 5 to Rule 1.4 addresses when lawyers must disclose AI use to clients.
The trigger is not the mere use of AI. Disclosure is required when a lawyer’s use of technology “presents a significant risk or materially affects the scope, cost, manner, or decision-making process of representation.” At that point, the lawyer must communicate “sufficient information regarding the use of technology to permit the client to make informed decisions regarding the representation.”
Two things to note. First, this is not a blanket disclosure requirement — more routine AI-assisted drafting or research may not cross the threshold in every matter. Second, the obligation is ongoing: lawyers must reassess their disclosure duty throughout a representation based on the novelty of the technology, the scope of the matter, and the sophistication of the client.
In practice, this means the client conversation about AI is not a one-time checkbox at the intake stage. It needs to be revisited as AI tools change and as representations evolve.
Rule 1.6 — Confidentiality: A New Definition of “Reveal”
The proposed amendment to Rule 1.6 expands what it means to “reveal” confidential client information in a way that directly addresses AI tools.
Under the proposed language, inputting client information into an AI system constitutes a revelation of confidential information if there is “a material risk that the information may be accessed, retained, or used, whether by the technological system or another user of that technological system, in a manner inconsistent with the lawyer’s duty of confidentiality.”
Attorneys accustomed to thinking about confidentiality as a prohibition on disclosing information to humans need to reconsider how they evaluate AI tools. A cloud-based AI system with an unclear or unfavorable data retention policy is not a neutral tool — under this proposed rule, using it with client data without proper vetting could itself be a breach.
This makes vendor due diligence a professional responsibility issue, not just a technology procurement question.
Rule 3.3 — Candor: Citations Are Your Problem
The proposed amendment to Rule 3.3 addresses AI hallucinations directly. A new comment states that “a lawyer’s duty of candor towards the tribunal includes the obligation to verify the accuracy and existence of cited authorities, including ensuring no cited authority is fabricated, misstated, or taken out of context, before submission to a tribunal, including any cited authorities generated or assisted by artificial intelligence or other technological tools.”
This codifies what courts have been saying in sanction orders for two years. AI-generated citations are not exempt from the existing prohibition on misquoting authority or citing overruled decisions. The verification obligation is the lawyer’s, not the software’s.
The five-category hallucination taxonomy we document throughout Ethics-Safe AI Use for Law Firms — fabricated authority, false quotation, misrepresented holding, outdated authority, and incomplete disclosure — maps directly onto the failure modes Rule 3.3 is now explicitly designed to catch. Each category represents a distinct verification step that the Mata Protocol is designed to surface before anything reaches a tribunal.
Rule 5.1 — Managerial Lawyers: AI Governance Is Now Your Responsibility
The proposed amendment to Rule 5.1 adds AI governance to the list of matters that managerial lawyers must address through firm policies and procedures.
The existing comment already references policies for conflicts, calendaring, and client funds. The new language adds that managerial lawyers must make reasonable efforts to establish procedures “governing the use of artificial intelligence, in accordance with the Rules of Professional Conduct.”
This lands directly on managing partners, practice group chairs, and any attorney who supervises other lawyers. “Aspirational statements” about responsible AI use do not satisfy this standard. Functioning governance — clear policies, training, and verification workflows — does.
For firms that have been taking a wait-and-see approach to AI policy, the wait is over.
Rule 5.3 — Supervision of Staff: AI Use Filters Down
The corresponding amendment to Rule 5.3 extends the AI obligation to the supervision of paralegals, legal assistants, law clerks, and all other nonlawyer staff who use AI tools in their work.
The existing comment requires that lawyers give nonlawyer assistants “appropriate instruction and supervision concerning all ethical aspects of their employment.” The proposed amendment adds “including the use of technology in the provision of legal services, such as artificial intelligence.”
Given how broadly AI tools have proliferated throughout law firm operations — document review, research assistance, correspondence drafting — this effectively means that every attorney who supervises staff needs to understand what AI tools those staff members are using and how they are being used.
The supervisory exposure is not hypothetical. If a paralegal submits a research memo built on an unverified AI output and an attorney incorporates it into a filing without independent review, both the supervisory failure and the competence failure are potentially at issue.
What This Means If You Practice in California Today
These rules are proposed, not final. COPRAC will review public comment and may modify them before they advance. The California Supreme Court has final authority over the Rules of Professional Conduct.
But the direction is unmistakable, and the underlying obligations are already implicit in the existing rules — which is precisely COPRAC’s point. Attorneys who are independently reviewing AI output, disclosing material AI use to clients, vetting their AI vendors for confidentiality compliance, verifying every citation before filing, and maintaining actual AI governance policies are already compliant with what these amendments require.
Attorneys who are not should not wait for the rules to become final.
The pattern documented across our AI-legal-research case database is consistent: sanctions do not require bad intent. They follow from the absence of verification. The attorneys disciplined over the past two years did not mean to submit fabricated citations — they relied on AI output without the independent review step the proposed Rule 1.1 now makes explicit.
California attorneys have until June 27 to get ahead of this. The Newport Beach AI Ethics Training Workshop on June 27 is a one-day intensive designed specifically for attorneys who want to understand their obligations under Rules 1.1, 1.6, 3.3, 5.1, and 5.3 — and leave with a working verification workflow they can implement immediately. Early-bird pricing at $399 ends May 29.
If you want to go deeper on the framework before the workshop, Ethics-Safe AI Use for Law Firms covers the seven Model Rules, the five-hallucination taxonomy, and the Mata Protocol — the tools the workshop is built around. Request a free chapter to start there.