25 AI Ethics Questions Florida Small-Firm Lawyers Are AskingA plain-English Florida Bar Opinion 24-1 AI ethics guide, plus the ABA opinion and the new court rule on AI citations.

Is it ethical for lawyers to use AI in Florida? Yes, under conditions the Florida Bar and the ABA have now spelled out in real detail. AI does not change who is responsible for the work. It changes what a competent review process has to catch. The 25 questions below are those conditions, not theory. The answers below are grounded in four authorities reviewed on September 12, 2026: The ABA opinion and the Florida Bar AI ethics opinion are advisory. The applicable professional-conduct rules, court rules, orders, and law control. Check every jurisdiction and court that applies to your work.

Need a quick answer? Use the 25 questions below. Need to put firm controls in place? Read Build your firm’s AI controls for the operating steps and 30-day plan.

Part One: Competence and Client Communication

1. Do I have a duty to understand AI?

Yes, at a practical level. Florida’s comment to Rule 4-1.1 now expressly says maintaining competence includes understanding the benefits and risks of technology, including generative AI. ABA Opinion 512 says a lawyer using a generative-AI tool needs a reasonable understanding of that tool’s capabilities and limits. You do not need to become a technologist. You do need enough understanding to decide whether, where, and how the tool may be used.

2. Am I required to use AI?

No. Competence does not require any particular tool or method. A lawyer may decide that a checklist, template, conventional research process, or human-only workflow is the better choice. The duty is to provide competent representation, not to adopt a product.

3. Can I charge a client for learning an AI tool?

Usually not for baseline education. ABA Opinion 512 says a lawyer may not charge a client to learn a tool the lawyer will regularly use for clients. Florida Opinion 24-1 similarly warns against charging for the time needed to develop minimal competence. Case-specific research or learning a particular tool at a client’s express request may be chargeable when reasonable and agreed in advance, preferably in writing.

4. What does a reasonable understanding of a tool include?

At minimum, know what the tool can and cannot reliably do; what happens to prompts, files, and outputs; who can access them; how long they are retained; whether they may be used for training or product improvement; what permissions or connected data the tool can reach; and how output will be verified. Read the current terms and privacy materials or get help from someone who can evaluate them. A marketing label is not a data map.

5. Do I have to tell clients I am using AI?

Sometimes. ABA Opinion 512 makes the answer fact-specific. Disclosure or consultation may be required when informed consent is needed to protect confidential information, the client asks, the engagement terms require it, AI use affects the fee, or the output will influence a significant decision in the representation. Routine use does not create one universal disclosure rule, and using AI in a draft does not by itself answer the question.

Part Two: Confidentiality and Data Handling

6. Can I paste client information into ChatGPT?

Do not put client information into any consumer AI account unless the firm has approved the exact account, settings, terms, retention, sharing, access, and use. OpenAI currently says content from individual services may be used for training unless the user opts out, while Temporary Chat is not used for training and business-product data is excluded from training by default. Those distinctions matter, but training is only one issue. Confidentiality also depends on retention, access, security, connected services, the sensitivity of the information, and whether client consent is required. Start with de-identified or minimum-necessary information. To check what a specific vendor does with your data, use the Vendor AI and Data Governance Audit and the AI Vendor Evaluation Worksheet.

7. What does “self-learning” mean in the Florida opinion?

Florida Opinion 24-1 uses the term for systems that may use additional inputs as they develop. The practical question is not the label. It is whether information is retained, shared, used to improve a model or service, or made accessible to other people or matters. Investigate the current terms and settings before entering confidential information. No. Consent depends on the information, the tool, the data flow, the safeguards, and the applicable rule. If confidential information will be disclosed to a third-party system in a way that requires consent under Rule 4-1.6, the consent must be informed. ABA Opinion 512 says general boilerplate is not enough when meaningful confidentiality risks require a real explanation. A use that does not disclose information relating to the representation may not require consent on confidentiality grounds.

9. Are Microsoft 365 Copilot and Gemini for Google Workspace safe for client work?

They publish stronger business-data protections than many consumer accounts. As of September 12, 2026, Microsoft says prompts, responses, and Microsoft Graph data used by Microsoft 365 Copilot are not used to train foundation models. Google says Gemini for Workspace submissions are not human-reviewed or used to train models outside the customer’s domain without permission. That does not create blanket approval. Confirm the exact license, feature, retention, permissions, connectors, web-grounding behavior, administrative settings, and contract before client information is used. For a plain-English look at each, see the Microsoft 365 Copilot guide and the Gemini and NotebookLM guide.

10. If the vendor calls a tool private, am I in the clear?

No. “Private” may describe one part of the data flow. It does not answer who can see the data, how long it is kept, which connected sources the tool can reach, whether users are oversharing through existing permissions, or whether the output is accurate. It also does not remove supervision, verification, communication, or fee duties.

Part Three: Accuracy, Research, and Court Filings

11. What is a hallucination?

It is plausible-sounding output with no reliable factual basis. It may be a nonexistent case, a false quotation, a misstated holding, an invented fact, or a real authority cited for a proposition it does not support. Generative systems produce probable text, not verified truth. Fluent writing is not evidence.

12. Have lawyers actually been sanctioned for inaccurate AI output?

Yes. Florida Opinion 24-1 cites Mata v. Avianca, where lawyers and their firm were sanctioned after false citations were filed. Florida courts have also addressed fabricated or inaccurate authorities. Effective June 15, 2026, Florida Rule of General Practice and Judicial Administration 2.515(d)(2) requires every signer of a filing to represent that identified legal authorities exist and are accurately cited. The rule expressly authorizes sanctions after notice and an opportunity to be heard.

13. If AI produces an error and I use it, who is responsible?

The lawyer. Florida Opinion 24-1 says lawyers remain responsible for their work product and professional judgment. ABA Opinion 512 reaches the same practical result through competence, supervision, meritorious-claim, candor, and communication duties. Software does not take the signature line. If AI-assisted work goes wrong, also check whether your malpractice policy has AI exclusions.

14. How should I verify AI output?

Return to the controlling source. Retrieve every cited case, statute, rule, quotation, and record reference. Confirm that it exists, is current, supports the stated proposition, and is quoted accurately. Check facts against the client file or another authoritative record. For Florida court filings, the signer’s Rule 2.515(d)(2) representation makes the authority check explicit. The review method should match the task and the risk, but no important proposition should be trusted because the prose sounds confident. Yes, as an aid. Use it to frame issues, generate search terms, organize possibilities, or work within a platform grounded in identifiable legal sources. Then retrieve and read the authority yourself. Confirm existence, jurisdiction, currency, treatment, quotations, and the proposition for which the authority is cited. A generated answer is a research lead, not the research file.

Part Four: Conflicts, Intake, and Chatbots

16. Can AI use create conflict problems?

It can create confidentiality and access-control problems that affect conflicts management. Information from one matter should not become available to people working on an adverse matter or be reused across matters without authorization. Review workspace permissions, shared histories, connected repositories, ethical walls, vendor access, and matter separation. Treat the system as part of the firm’s information environment, not as a private notebook by default.

17. What rules apply to an intake chatbot?

Florida Opinion 24-1 says a generative-AI chatbot must identify itself as an AI program, not a lawyer or law-firm employee. It should not provide legal advice, and legal questions should go to a lawyer. The firm remains responsible for misleading, intrusive, or coercive communications. Clear warnings should limit the firm’s obligations, and the intake design should minimize substantive facts until the firm has enough information to run an appropriate conflict process.

18. Can information sent through a chatbot create prospective-client duties?

Possibly. Florida Opinion 24-1 points to Rule 4-1.18 and explains that electronic communications may amount to a consultation depending on the circumstances. A site that invites matter details without clear, understandable limits creates more risk than a site that provides general information and controlled contact options. Design the intake flow before turning on the chatbot.

Part Five: Fees and Supervision

19. If AI saves time, can I bill the hours the task used to take?

Not under an hourly arrangement. ABA Opinion 512 says hourly billing must reflect time actually spent, including reasonable prompting, review, revision, and professional judgment. Florida Opinion 24-1 prohibits duplicate charges and falsely inflated time. Flat and contingent fees are not measured by the clock in the same way, but they still must satisfy the applicable reasonableness requirements. Charges for an AI service must also be disclosed and handled consistently with the rules on costs and overhead.

20. Should I consider flat fees when AI compresses the work?

Maybe. Florida Opinion 24-1 identifies flat or contingent arrangements as alternatives that can let efficiency benefit both lawyer and client. The right structure depends on the matter, risk, predictability, client expectations, and governing fee rules. AI does not excuse a clearly excessive or otherwise improper fee.

21. What must supervising lawyers do when staff or associates use AI?

Set clear rules, train the people using the tool, and review the work. ABA Opinion 512 says managerial lawyers must establish clear policies for permissible use and supervisory lawyers must make reasonable efforts to ensure that lawyers and nonlawyers comply with professional duties. Florida Opinion 24-1 requires policies and practices that make AI use consistent with the lawyer’s obligations. Vendor use and outsourced work need supervision too.

22. Does the firm need a written AI policy?

A short written policy is the clearest practical way to implement the duties, even though the authorities do not prescribe a particular page count or template. It should identify approved tools and accounts, prohibited uses, information-handling rules, verification steps, disclosure and consent triggers, supervision, incident escalation, and the person who answers questions. A policy nobody can find or apply is not much of a policy. The Free Florida AI Starter Playbook includes client disclosure language and a tool-approval scorecard.

Part Six: Advertising and Vendor Claims

23. Can a law firm advertise that it uses AI?

Yes, if the statement complies with the advertising rules. Florida Opinion 24-1 says a firm may advertise its use of generative AI but may not claim its AI is superior to another lawyer’s or firm’s system unless the claim is objectively verifiable. The firm is also responsible for misleading chatbot statements and must identify an AI chatbot as an AI program rather than a lawyer or employee.

24. What should I ask an AI vendor before approving the tool?

Ask for the current contract and written answers about confidentiality, security, retention, deletion, model training, product improvement, human access, subprocessors, incident notice, service-of-process notice, data location, permissions, connected systems, export, termination, and liability. Ask how the system supports matter separation and verification. Security certifications may be useful evidence, but no badge is an ethics safe harbor. The AI Vendor Evaluation Worksheet and the Vendor AI and Data Governance Audit walk through this review.

25. Where should a small firm start?

Start with the work, not the product.
  1. Inventory the exact AI accounts, licenses, settings, and unsanctioned uses already present.
  2. Pick one recurring, lower-risk workflow with a visible cost or delay.
  3. Decide which information may enter which approved system.
  4. Name the lawyer or staff member who checks each output, and define the check.
  5. Write the process down, train the users, and measure whether the workflow improved.
AI use is optional. Competence, confidentiality, accuracy, supervision, candor, communication, and reasonable fees are not. Small firms do not need AI theater. They need fewer dropped balls, and a process that catches the ones AI drops too. If you are not sure where your firm stands today, the 2-minute readiness self-check is a free first step. If you want a second set of eyes on your firm’s tools and workflow, schedule a 15-minute introductory call. We will identify the question, the current process, and the smallest sensible next step. Greg Yates is the founder of Actionable AI and a former Big Law equity partner. He helps Florida law firms with fewer than 25 lawyers improve recurring work using tools they may already pay for. This page provides general educational information, not legal advice. Check the current professional-conduct rules, ethics opinions, court rules, standing orders, engagement terms, and other authority that apply to your jurisdiction and matter.