AI Ethics Checklist for Solo Lawyers: What the Bar Opinions Actually Require

We may earn money or products from the companies mentioned in this post.

Disclosure: This article contains affiliate links. If you sign up for a product through one of these links, AI for Law Firms may earn a commission at no extra cost to you.

Bar guidance on generative AI usually comes down to two sentences: be competent and protect confidentiality. That’s accurate, but a solo can’t do much with it on a Monday morning. This checklist turns the main opinions into specific steps: ABA Formal Opinion 512, Texas Opinion 705, Florida Opinion 24-1, Oregon Formal Opinion 2026-208, California’s practical guidance, and Virginia LEO 1901. Each item links to the source so you can read it yourself.

Not legal advice. This is general information about published ethics guidance as of October 2026. Ethics opinions are advisory, they change, and your own jurisdiction’s rules and opinions control. For an overview by state, see our AI ethics rules by state hub.

Why AI raises new ethics questions

The Model Rules don’t depend on any particular technology. What’s new is the failure mode. A research database either finds a case or it doesn’t. A generative model predicts plausible text, so it can produce a citation with the right format, a real court name, and a believable year for a case that doesn’t exist. That’s why every opinion below spends so much time on verification.

The Texas committee put it bluntly: “lawyers are responsible for the work product they submit regardless of who (or what) does the original research and drafting” (Texas Opinion 705).

The checklist

1. Competence (Rule 1.1): understand the tool before you use it on a matter

☐ Know whether the tool can invent authority. (General chatbots can.)
☐ Know whether it searches a maintained legal database or only generates from training data.
☐ Test it on a question you already know the answer to before you rely on it for that kind of task.
☐ Know its weak spots: poor scans, very long documents, tables, handwriting.

Texas Opinion 705 says a lawyer who uses generative AI “must have a reasonable and current understanding of the technology.” It adds that Rule 1.01 “almost certainly does not require the use of generative AI,” but lawyers shouldn’t “unnecessarily retreat” from tools that save clients money.

2. Confidentiality (Rule 1.6): know where the data goes

☐ Read the current terms for training, retention, and human access.
☐ Use firm-managed business plans, not personal accounts, for any client information.
☐ Turn off memory and shared links for client work.
☐ Strip identifying fact combinations, not just names.

Florida Opinion 24-1 says lawyers must protect confidentiality “by researching the program’s policies on data retention, data sharing, and self-learning” (Florida Bar Ethics Opinion 24-1). Texas lists concrete precautions: understand how the technology works, review and possibly renegotiate the terms of service, learn the data-security protections, and train staff. For a plan-by-plan breakdown of what OpenAI, Anthropic, and Microsoft promise, see our guide to AI tools and client data. Our redaction checklist covers the mechanics. For PDFs and scans, a tool like Redactable can automate much of it.

3. Informed consent: boilerplate won’t do it

☐ Before you put client information into a self-learning tool, get informed consent.
☐ Make the consent specific: which tool, what information, the risks, the benefits, and the client’s right to decline.
☐ Write it down, even if your rules don’t strictly require writing.

ABA Formal Opinion 512 says: “To obtain informed consent when using a GAI tool, merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient.” It also says the client must get “the lawyer’s best judgment about why the GAI tool is being used, the extent of and specific information about the risk… and a clear explanation of the GAI tool’s benefits to the representation.” Florida “recommends” informed consent before using a third-party program that would involve disclosing confidential information. Texas says lawyers “should consider informing clients” and “may need to secure client consent.”

4. Communication (Rule 1.4): tell clients when AI use is material

☐ Disclose AI use when it materially affects the representation, its cost, or its risk.
☐ Routine uses (spell-check, formatting) generally don’t need a separate discussion.
☐ For substantive uses, a sentence in a status update is a sensible default.

5. Candor and verification (Rules 3.1, 3.3): open every source

☐ Open and read every cited authority in an authoritative source.
☐ Confirm each quotation word for word.
☐ Check subsequent history with a citator.
☐ Confirm jurisdiction, court level, and procedural posture.
☐ Don’t use a second AI tool as your only check on the first.
☐ Check local rules, standing orders, and the judge’s page for any AI certification requirement before you file.

Florida and Texas both cite Mata v. Avianca (S.D.N.Y. 2023), where lawyers were sanctioned for citing cases ChatGPT invented. Florida notes that failing to verify can violate the duties of competence, meritorious claims, candor, and truthfulness. Texas points out that some courts require AI disclosure (it cites N.D. Tex. LR 7.2(f)) while others have declined to adopt special rules. See our citation verification workflow.

6. Supervision (Rules 5.1, 5.3): vendors and contractors count

☐ Treat an AI vendor like a nonlawyer assistant: do due diligence before you use it, and oversee it afterward.
☐ Give contract paralegals and virtual assistants a written list of approved tools.
☐ Spell out what information may and may not be entered.

Florida says the duty to supervise applies to nonlawyers “both within and outside of the law firm” and that a third-party operator “does not obviate the need” to ensure compliance. Oregon Formal Opinion 2026-208, on chatbots and AI agents, treats the chatbot vendor as subject to RPC 5.3 supervision.

7. Fees (Rule 1.5): bill real time, and explain flat fees

☐ When billing hourly, bill only time actually spent: prompting, reviewing, verifying, and revising.
☐ Don’t bill the hours the AI saved you.
☐ Don’t bill clients for learning a general AI tool.
☐ Pass through actual per-use AI costs only with the client’s advance agreement. Treat flat subscriptions as overhead.
☐ For flat fees, make sure the fee stays reasonable and explain its basis.

Texas: “A lawyer may not, however, charge hourly fees for the time that was ‘saved’ by using the generative AI program.” Florida: efficiency “must not result in falsely inflated claims of time,” and a lawyer “should be careful not to charge for the time spent developing minimal competence in the use of generative AI.” Virginia LEO 1901, approved by the Supreme Court of Virginia effective November 24, 2025, leaves more room for non-hourly fees. It holds that reduced time does not by itself require a proportionately reduced flat fee, as long as the fee is reasonable and explained (Virginia LEO 1901). Our upcoming guide to billing for AI time savings covers the details.

8. Advertising and intake bots (Rules 7.1, 1.18): say it’s AI

☐ Any chatbot or AI receptionist should say it is AI and not a lawyer.
☐ It should not give legal advice or suggest that an attorney-client relationship has formed.
☐ Enter prospective-client information into your conflicts system promptly.

Florida: generative AI chatbots “must include a disclaimer indicating that the chatbot is an AI program and not a lawyer or employee of the law firm.” The Oregon opinion, in the version presented to the Board of Governors, notes that “more than one state has opined that the use of a chatbot that does not identify itself as a chatbot rather than a person could be a misleading communication” (OSB Board of Governors materials, Opinion 2026-208). See also what intake you can (and can’t) automate.

A sample one-page AI policy for a solo

Approved tools: list two or three by name and plan tier.
No-client-data tier: general research questions, templates, and plain-English explanations. Any approved tool may be used.
Client-data tier: only firm-managed business plans with no training by default, and only with client consent where required.
Never: material under a protective order that bars third-party processing, or any personal-account use for client work.
Verification: every authority is opened and read before it is used.
Billing: actual time only for hourly work. AI subscriptions are overhead.
Records: note which tool was used on which matter, and what was verified.

For a fuller version, adapt our ABA Formal Opinion 512 policy template.

What about California?

The State Bar of California published Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law in November 2023. It covers confidentiality, competence, supervision, and fees, and Texas cites it for the recommendation to tell clients when generative AI will be used. Read the document itself for California-specific wording, especially on billing.

Malpractice and insurance

Ethics compliance and malpractice coverage are separate questions. Carriers increasingly ask about AI use at renewal, so a written policy like the one above helps there too. See whether malpractice insurance covers AI mistakes.

Your first Monday: a 30-minute start

Ten minutes: list every AI tool you or anyone working for you has used in the last month, including browser extensions and meeting notetakers.
Ten minutes: for each tool, write down the plan tier and what its current terms say about training and retention. Drop or downgrade any tool you can’t answer for.
Five minutes: add a matter-specific AI consent paragraph to your engagement-letter workflow. Treat it as a conversation prompt, not boilerplate.
Five minutes: save the verification checklist above where you draft, and check your main courts’ standing orders for AI requirements.

Do this again every quarter. Terms and opinions change faster than most CLE cycles.

If a meeting notetaker turns up on your list, especially a browser extension or a free personal account, replace it with a tool you can actually vet. Sembly AI is one option for solos who want meeting summaries and action items from Zoom, Google Meet, Microsoft Teams, or Webex calls. Run it through the same tier-and-terms check above, review its security and data terms, and get recording consent where required before using it on client calls.

Frequently asked questions

Does ABA Formal Opinion 512 apply to solos?

Yes. It interprets the Model Rules, which apply to every lawyer. It isn’t binding in itself, but many states follow it or reach similar conclusions.

Can I use ChatGPT if I strip out names?

Removing names helps, but combinations of facts can still identify a client. Texas and Florida both focus on whether the tool could reveal confidential information. If it’s a consumer tool that may train on your inputs, keep client facts out.

Do I have to tell the court I used AI?

It depends on the court and the judge. Some require a certification or disclosure, and others don’t. Check local rules, standing orders, and the judge’s procedures before every filing.

Can I bill for time AI saved me?

Not on an hourly basis. Texas and Florida are explicit. For flat fees, Virginia LEO 1901 says the same fee isn’t automatically unreasonable just because AI made the work faster, but the fee still has to be reasonable and explained.

Key takeaways

Understand the tool, including whether it can invent authority, before you use it on a matter.
Client data goes only into firm-managed tools that don’t train on it, with informed consent where required. Boilerplate isn’t enough.
Verify every citation and quotation in an authoritative source.
Bill real time on hourly matters, and make sure any flat fee is reasonable and explained.
Chatbots and AI receptionists must identify themselves as AI.