Florida Bar Opinion 24-1 on Generative AI, Explained for Small Firms

The Florida Bar was among the first state bars to address generative artificial intelligence head on, and the result, Ethics Opinion 24-1, dated January 19, 2024, remains one of the more practical road maps a small firm can follow. It does not ban the technology. It permits lawyers to use generative AI while holding four familiar duties firmly in place: confidentiality, oversight, reasonable fees, and honest advertising.

This explainer walks through what Florida Bar Opinion 24-1 actually says, the rules it leans on, and what a firm of roughly three to thirty attorneys should do about it. The citations are included so you can read the source yourself.

What Opinion 24-1 is, and what it is not

Opinion 24-1 is an advisory ethics opinion issued by The Florida Bar's Board Review Committee on Professional Ethics at the direction of the Board of Governors. The opinion states plainly on its face that "Advisory ethics opinions are not binding." It is guidance, not a disciplinary rule. That said, it interprets the Rules Regulating The Florida Bar, and it reads those rules the way a grievance committee likely would, so treating it as persuasive is the prudent course.

The opinion's one-paragraph summary is the whole framework in miniature: lawyers may use generative AI but must "protect the confidentiality of client information, provide accurate and competent services, avoid improper billing practices, and comply with applicable restrictions on lawyer advertising." It expressly ties itself to the arrival of consumer generative AI, noting that the release of ChatGPT in late 2022 prompted the debate, and it grounds its competence discussion in the duty to understand "the benefits and risks associated with the use of technology" found in the comment to Rule 4-1.1.

A useful anchor for the whole discussion is the case the opinion cites as the cautionary tale: Mata v. Avianca, No. 22-cv-1461, 2023 WL 4114965 (S.D.N.Y. June 22, 2023), where a federal judge sanctioned two lawyers and their law firm after they filed a brief containing fake citations generated by an AI tool. The opinion is careful to quote the same court's observation that "there is nothing inherently improper about using a reliable artificial intelligence tool for assistance." The lesson is not abstinence. It is verification.

Confidentiality and client consent before third-party tools

The center of gravity in Florida lawyer AI ethics is confidentiality under Rule 4-1.6. The duty is broad, covering "all information learned during a client's representation, regardless of its source," and absent informed consent or an exception, a lawyer may not reveal it.

The headline takeaway is a consent recommendation, not an absolute mandate. The opinion says it is "recommended that a lawyer obtain the affected client's informed consent prior to utilizing a third-party generative AI program if the utilization would involve the disclosure of any confidential information." Two qualifiers matter here. First, the trigger is disclosure of confidential information to a third party. Second, the opinion draws a clean line for in-house tools: "If the use of a generative AI program does not involve the disclosure of confidential information to a third-party, a lawyer is not required to obtain a client's informed consent pursuant to Rule 4-1.6."

That distinction is the most consequential practical point in the opinion. A self-contained or properly walled-off tool that does not send client data outside the firm sits in a different posture than a public consumer chatbot. The opinion is also specific about the diligence the technology requires. A lawyer must understand whether a program is "self-learning," because a self-learning model "raises the possibility that a client's information may be stored within the program and revealed in response to future inquiries by third parties." It instructs lawyers to research a program's policies on data retention, data sharing, and self-learning.

Rather than invent a new standard, Opinion 24-1 borrows from the Bar's prior technology opinions. It cites Florida Ethics Opinion 12-3 on cloud computing (which in turn relied on New York State Bar Ethics Opinion 842 and Iowa Ethics Opinion 11-01) for a vendor-diligence checklist: confirm the provider is obligated to preserve confidentiality, will notify you of a breach or a subpoena, and disclose whether it retains your data after the engagement ends. It also leans on Florida Ethics Opinion 07-2 (overseas paralegal outsourcing) for the principle that a provider should get only the information necessary for the specific matter, and on Opinion 06-2 for the point that you should not try to pull confidential data that another lawyer fed into a shared tool.

Oversight and verification: treat AI like a nonlawyer assistant

Opinion 24-1's central analogy is that generative AI should be supervised the way a firm supervises a paralegal or other nonlawyer assistant. It acknowledges that Rule 4-5.3 defines a nonlawyer assistant as "a person," but it concludes that the supervisory standards "provide useful guidance" for AI use anyway.

Three duties follow directly. First, the firm must have policies reasonably assuring that the AI's output is compatible with the lawyer's professional obligations. The opinion warns that lawyers who lean on AI "risk many of the same perils as those who have relied on inexperienced or overconfident nonlawyer assistants." Second, the lawyer must review the work product. In the opinion's words, "a lawyer must verify the accuracy and sufficiency of all research performed by generative AI." Third, the duty does not disappear because a third party runs the tool; citing ABA Formal Opinion 498 (2021), the opinion confirms that supervisory duties reach nonlawyers "both within and outside of the law firm."

The opinion spells out the stakes of skipping verification. Failure to check AI output can lead to violations of the duties of competence (Rule 4-1.1), avoidance of frivolous claims (Rule 4-3.1), candor to the tribunal (Rule 4-3.3), and truthfulness to others (Rule 4-4.1), "in addition to sanctions that may be imposed by a tribunal," a direct echo of Mata v. Avianca. It also draws a hard line on delegation: a lawyer "may not delegate to generative AI any act that could constitute the practice of law such as the negotiation of claims or any other function that requires a lawyer's personal judgment."

Fees and costs: efficiency cannot inflate the bill

On billing, Opinion 24-1 applies long-standing fee principles to a new line item. Rule 4-1.5(a) bars illegal or clearly excessive fees and costs, and the opinion translates that into several concrete rules for AI.

• Charge actual cost, not a markup. A lawyer "may only ethically charge a client for the actual costs incurred on the individual client's behalf and must not duplicate charges that are already accounted for in the lawyer's overhead." The opinion cites The Florida Bar v. Carlon, 820 So. 2d 891, 899 (Fla. 2002), where a lawyer was sanctioned for, among other things, a flat $500 administrative charge applied to each client's file, and ABA Formal Opinion 93-379 (1993), for the principle that a lawyer should bill only costs that reasonably reflect actual costs.
• Do not prorate what you cannot attribute. If a lawyer cannot determine the cost tied to a particular matter, the opinion says the lawyer "may not ethically prorate the periodic charges of the generative AI and instead should account for those charges as overhead." A flat monthly software subscription generally belongs in overhead, not on the client's bill.
• Efficiency belongs to the client too. Because AI can compress hours of work, "this increase in efficiency must not result in falsely inflated claims of time." You cannot bill the time the work would have taken without the tool. The opinion suggests that flat fees or contingent arrangements can let "the benefits of increased efficiency accrue to the lawyer and client alike."
• Do not bill for learning the tool. Citing the principle in Attorney Grievance Commission of Maryland v. Manger, 913 A.2d 1 (Md. 2006), that general education should not be charged to a client, the opinion says a lawyer "should be careful not to charge for the time spent developing minimal competence in the use of generative AI."

Double-billing gets its own mention in the summary, which states flatly that using AI "does not permit a lawyer to engage in improper billing practices such as double-billing." For contingent-fee matters, the opinion reinforces Opinion 07-2's caution that it is improper to bill separately for work that is normally folded into the standard fee.

Chatbots and advertising: the tool must not mislead

The fourth area is where Opinion 24-1 is most specific, and it is the one many small firms overlook because a website chatbot feels like marketing rather than legal work. The opinion treats client-facing AI as both an advertising matter and a potential unauthorized-practice and prospective-client trap.

The bright-line requirement is disclosure. Generative AI chatbots that communicate with clients or third parties "must include a disclaimer indicating that the chatbot is an AI program and not a lawyer or employee of the law firm." A lawyer "must inform prospective clients that they are communicating with an AI program." This connects to Rule 4-7.13(b)(5), which addresses voices or images that create the false impression that the speaker is a lawyer or firm employee, and to Rule 4-7.13(b)(3), which limits claims about a lawyer's skills, experience, or record to those that are "objectively verifiable." Practically, you may advertise that your firm uses AI, but you cannot claim your AI is superior to others' unless that claim is objectively verifiable.

The opinion also flags two relationship risks. Drawing on Florida Ethics Opinion 88-6 (nonlawyers conducting initial interviews) and the comment to Rule 4-1.18, it warns that an "overly welcoming" chatbot can inadvertently create a prospective-client or even a lawyer-client relationship, since that relationship can turn on the person's reasonable belief, citing Bartholomew v. Bartholomew, 611 So. 2d 85 (Fla. 2d DCA 1992). A chatbot that dispenses anything resembling legal advice is the danger. The opinion's suggested guardrails: a clear AI disclaimer, cautionary statements that limit the lawyer's obligations, and "screening questions that limit the chatbot's communications if a person is already represented by another lawyer."

A practical checklist for a 3-to-30-lawyer firm

Opinion 24-1 maps cleanly onto a short set of firm actions:

• Pick the tool deliberately. Prefer tools that do not train on your inputs and that contractually protect confidentiality; document the vendor's retention, sharing, and self-learning policies before you adopt it.
• Get consent when data leaves the firm. If a third-party tool will receive confidential information, obtain the client's informed consent, preferably in writing; for in-house tools that do not disclose to a third party, consent is not required under Rule 4-1.6.
• Verify every output. Treat AI like a new paralegal: check every citation and factual assertion before anything reaches a client or a court, and never delegate judgment calls or the practice of law.
• Bill honestly. Pass through only actual, attributable costs; place flat subscriptions in overhead; do not bill phantom hours saved or time spent learning the tool.
• Fix your chatbot. Add a conspicuous "this is an AI, not a lawyer" disclaimer, screen for represented persons, and make sure it never gives legal advice.
• Write it down. Adopt a brief written AI policy so the supervisory and confidentiality duties are demonstrably in place, not improvised.

None of this is exotic. Opinion 24-1's own closing note is that these duties are "still in its infancy" and "should not be treated as an exhaustive list." The throughline is continuity: generative AI is a new instrument played under old rules, and the firms that do well with it will be the ones that keep the lawyer, not the model, accountable for the result.

This is general information for lawyers and law-firm leaders, not legal advice, and it does not create an attorney-client relationship. The authorities are cited so you can read them yourself.