Skip to content
Zrico
Menu

For law firms

State Bar AI Ethics Opinions in Plain English

California, Florida 24-1, New York, Texas 705, Pennsylvania 2024-200, New Jersey and more: what each says on AI, and how it differs from ABA Opinion 512.

Updated 8 min read

Since late 2023, state bars and courts have published guidance on how lawyers may use generative AI. None of it creates new rules. Each document explains how the rules you already follow apply. This guide summarizes the major ones as of October 2026, with each linked to the issuing body’s own page or PDF, and then compares them with ABA Formal Opinion 512.

A few points before the table. Ethics opinions are advisory: Florida’s says plainly that its opinions are not binding. California’s document is practical guidance rather than a formal opinion. New Jersey’s comes from its Supreme Court, which regulates lawyers there, but it says it does not change the Rules of Professional Conduct. Always check your own state.

At a glance

Jurisdiction Document Date Client consent for confidential information Billing
California State Bar Practical Guidance Nov. 16, 2023; revised May 14, 2026 Required before using a tool with material confidentiality or security risks Actual time; general AI subscriptions are overhead; no markup on AI costs without informed written consent
Florida Ethics Opinion 24-1 Jan. 19, 2024 Recommended before a third-party tool receives confidential information No inflated hours; actual costs only, disclosed in advance
New York State Bar Task Force on AI report and guidelines Approved Apr. 6, 2024 Get provider assurances even when the client consents Any surcharge above cost must be stated in the engagement letter
New York City Bar Formal Opinion 2024-5 Aug. 7, 2024 Required before input into a tool that shares information with third parties No hourly charge for time AI saved
Texas Opinion 705 Feb. 2025 If not reasonably satisfied the tool is safe, consult the client and get consent Bill time spent, not time saved; costs only if the client agrees
Pennsylvania Joint Formal Opinion 2024-200 (PBA and Philadelphia Bar) 2024 Educate the client and get consent where needed AI expenses must be reasonable and disclosed
New Jersey Supreme Court Preliminary Guidelines Jan. 24, 2024 Ensure the tool is secure before entering non-public information Left for future guidance
District of Columbia Ethics Opinion 388 Apr. 2024 Use secure tools or keep confidential information out Hourly: actual time only
Kentucky Ethics Opinion KBA E-457 Mar. 15, 2024 Continuing duty to safeguard information Consider reducing fees; AI expenses only with advance written agreement
North Carolina 2024 Formal Ethics Opinion 1 Nov. 1, 2024 Avoid client-specific information in public tools Actual time; expenses with client consent

The opinions, briefly

California. The State Bar’s Committee on Professional Responsibility and Conduct wrote the original guidance in November 2023. The Board of Trustees approved a revised version on May 14, 2026, which replaces it and adds agentic AI, meaning tools that act on their own. A lawyer must not input confidential client information into a tool that presents material confidentiality or security risks without informed consent, and relying on general marketing assurances is not enough. Lawyers “must consider” telling clients they intend to use AI. Agentic tools must not file documents or communicate with a court without lawyer review.

Florida. Opinion 24-1 lets lawyers use generative AI if they protect confidentiality, check accuracy, bill properly and follow advertising rules. It recommends informed consent before using a third-party tool that would receive confidential information, and says consent is not needed if no third party receives it. A firm may not prorate a subscription across clients; if it cannot find the actual cost for a matter, the cost is overhead. Intake chatbots must tell people they are talking to an AI program, not a lawyer or employee.

New York. The New York State Bar Association’s Task Force report, approved by its House of Delegates on April 6, 2024, includes a guidelines chart. It suggests an engagement letter statement that AI may be used, with the client’s acknowledgment, and says that even with consent you should get assurance the provider will protect and segregate client information. The New York City Bar’s Formal Opinion 2024-5 says lawyers must not put confidential information into a tool that shares it with third parties without informed consent. Closed systems inside the firm avoid that risk, but ethical walls still apply. Lawyers should consider telling clients about AI that is not routinely used.

Texas. Opinion 705 says competence almost certainly does not require using AI, but a lawyer who uses it must understand it. If the lawyer is not reasonably satisfied a tool will protect confidential information, the lawyer should at minimum not enter it without consulting the client and getting consent. Lawyers may bill time actually spent, including reviewing output, but not time saved.

Pennsylvania. The Pennsylvania Bar Association and Philadelphia Bar Association issued Joint Formal Opinion 2024-200 together. It says a lawyer using AI must understand the tool, check every citation and the material cited, and educate the client and seek informed consent. AI-related expenses must be reasonable and disclosed.

New Jersey. The Supreme Court’s preliminary guidelines, dated January 24, 2024, say a lawyer must check and verify all AI output. There is no general duty to tell clients each time AI is used, but a lawyer must tell a client who asks or who needs the information to make an informed decision. A lawyer must ensure a tool is secure before entering non-public client information.

Other states. D.C. Opinion 388 notes that free consumer tools often train on inputs and that anonymizing may not be enough. Kentucky E-457 says there is no duty to disclose routine AI use unless the work is outsourced, AI costs are charged, or a court rule requires it, and AI training costs may not be billed. North Carolina 2024 FEO 1 says to avoid client-specific information in public tools, and that substantive work delegated to AI may need consent, like outsourcing.

How they differ from ABA Formal Opinion 512

ABA Formal Opinion 512, issued July 29, 2024, is the national reference point. The state documents mostly agree with it, with some real differences.

Consent. Opinion 512 requires informed consent before you put client information into a self-learning tool, and says boilerplate in an engagement letter is not enough. Florida only recommends consent. Texas uses a “reasonably satisfied” test. New Jersey and Kentucky focus on security rather than consent. California and the New York City Bar tie consent to the tool’s risks and data sharing, not to whether it learns from inputs. The New York State Bar’s engagement letter acknowledgment may help with disclosure, but under Opinion 512 it would not replace informed consent for a self-learning tool.

Telling clients. Opinion 512 lists situations where you must disclose AI use. New Jersey and Kentucky say there is no general duty. California says you must consider it.

Fees. All agree you bill actual time. California’s 2026 revision adds that AI costs may not be marked up without the client’s informed written consent. Florida bars prorating subscriptions. The New York State Bar allows a surcharge above cost if the engagement letter states it and the total stays reasonable.

Topics 512 does not cover. Florida addresses chatbot advertising and intake. California’s 2026 revision addresses agentic AI.

What to do

  1. Read your own state’s guidance in full, plus Opinion 512.
  2. Where they differ, follow the stricter rule, which is usually informed consent before confidential information goes into a tool you cannot fully vouch for.
  3. Read each tool’s terms on training, retention and disclosure, and record what you found.
  4. Put AI use, consent and billing terms in your engagement letter, and do not rely on it alone for consent.
  5. Check for updates twice a year. California has already revised its guidance once.

How Zrico fits

Most of this guidance asks you to know where client information goes and who can see it. Zrico gives each firm its own private workspace, with its own application and databases. Claude is the default and is processed only in the US. GPT and Gemini are off unless the firm owner turns them on, and their processing is not limited to the US. No AI provider (Anthropic, OpenAI or Google) trains on prompts, files or answers. Sign-in is invite-only with two-factor authentication, required for admins, and the firm has a usage and audit report and retention settings to support supervision. Microsoft 365 sign-in and client-name masking are coming soon. Zrico has no SOC 2, ISO or HIPAA certification, does not make a firm compliant with any opinion, and every output must be reviewed by a lawyer.

Sources

General information, not legal advice. Check the current rules in your jurisdiction.

Give your firm a private place to use AI.

$39 per seat per month after a 30-day free trial, 5-seat minimum. Your workspace is ready minutes after you sign up.

Your card is charged when the trial ends. Cancel any time before then and you pay nothing.

Start 30-day free trial