ABA Formal Opinion 512: A Checklist for Small Firms

The ABA issued Formal Opinion 512 on July 29, 2024. It was the ABA’s first formal ethics opinion on generative AI, and it does not create new rules. It explains how the Model Rules you already follow apply when you use AI tools, and this guide turns it into steps a small firm can take.

What the opinion covers

Opinion 512 looks at generative AI tools: software that writes text, summarizes documents or drafts content in response to a prompt. It covers general-purpose tools and tools built for lawyers. The committee notes that these tools change fast and that its guidance may change with them.

ABA opinions are not binding on their own. Your state’s rules of professional conduct, its ethics opinions and your local court rules control. Opinion 512 itself cites AI guidance from several state bars, so read it alongside your own state’s opinions.

Competence (Rule 1.1)

You do not need to become an AI expert. You do need a reasonable understanding of what the specific tool you use can and cannot do. You can get there by reading, by CLE, or by drawing on someone who knows the tool. The opinion says this is an ongoing duty, because the tools keep changing.

The core risk is inaccurate output. AI tools can produce plausible text with no basis in fact, which the opinion calls hallucinations. Relying on output without an appropriate degree of independent review could breach Rule 1.1.

How much review is enough depends on the tool and the task. The opinion gives an example: if you tested a tool’s contract summaries against a sample you reviewed yourself and found them accurate, you may not need to check every summary by hand. But you cannot hand off professional judgment. A tool cannot give legal advice to a client or negotiate a claim on its own. You remain fully responsible for the work.

Confidentiality (Rules 1.6, 1.9(c) and 1.18(b))

Before you enter information relating to a representation into an AI tool, you must weigh the risk that people outside the firm, or people inside the firm who should not see it, could access it. The same duty covers former and prospective clients.

The opinion singles out “self-learning” tools: tools that learn from what users put into them. With these tools, one client’s information could surface in output that another lawyer uses on a different matter. Because of that risk, the opinion concludes that you need the client’s informed consent before you input information relating to the representation into such a tool.

Informed consent has real content. The client needs your explanation of why you want to use the tool, the specific risks, what kinds of information would go in, how others could use it against the client, and what the tool adds. The opinion states that general, boilerplate language in an engagement letter is not sufficient.

As a baseline, the opinion says lawyers should read and understand the tool’s terms of use, privacy policy and related contract terms, or rely on a colleague or outside expert who has. If you use a tool only for idea generation and enter no client information, client consent is not needed for that use.

Communication (Rule 1.4)

Even where Rule 1.6 does not require consent, Rule 1.4 may require you to tell the client about AI use. The opinion lists situations where you must disclose:

• The client asks how you did the work or whether you used AI.
• The engagement agreement or the client’s outside counsel guidelines require it.
• You plan to input information relating to the representation, as above.
• AI use bears on the basis or reasonableness of your fee.
• AI output will influence a significant decision, such as evaluating likely litigation outcomes or jury selection.

Outside those cases, disclosure depends on the facts. The opinion suggests the engagement agreement is a sensible place to explain how the firm uses AI.

Meritorious claims and candor (Rules 3.1, 3.3 and 8.4(c))

Rule 3.1 bars frivolous claims and contentions. Rule 3.3 bars false statements of law or fact to a tribunal. The opinion notes that even an unintentional misstatement to a court can be a misrepresentation under Rule 8.4(c).

Before you file anything, review the AI-assisted material, including analysis and citations, and fix errors. The opinion names nonexistent case citations, wrong analysis of authority and misleading arguments as problems that have already happened. Some courts now require lawyers to disclose AI use, so check the local rules and any standing orders of your judge.

Supervision (Rules 5.1 and 5.3)

Lawyers with managerial authority must set clear policies on the firm’s permitted use of AI. Supervising lawyers must make reasonable efforts to ensure that lawyers and staff follow them. That includes training on how the tools work, their limits, the ethics issues and secure data handling.

Rule 5.3 also reaches outside providers. The opinion applies earlier guidance on outsourcing and cloud computing to AI vendors. It suggests you check that the tool is configured to protect confidentiality, that the vendor’s obligation is enforceable, that you will be told about a breach or a demand for client data, what the vendor retains during and after the relationship, and how reliable and secure its service is.

Fees (Rule 1.5)

If you bill hourly, bill for the time you actually spent. If a tool lets you draft a pleading in 15 minutes, you bill the 15 minutes plus your review time, not the time the task used to take.

For flat or contingent fees, the Rule 1.5(a) factors still apply. The opinion says it may be unreasonable to charge the same flat fee with AI as without it if the tool makes the work much faster.

Treat general-purpose AI that works like ordinary office equipment as overhead. Do not pass that cost to clients unless you disclosed it in advance. A per-use charge from a third-party tool used for one client’s matter can ordinarily be billed at actual cost. And you may not bill a client for time spent learning a tool you will use regularly, unless the client specifically asked you to use a tool you did not know.

The checklist

1. Name an owner. Pick one partner to be responsible for AI decisions and the firm’s AI policy.
2. List every AI tool in use. Include free tools staff use on their own, and AI features built into software you already pay for.
3. Read the terms for each tool. Record whether it trains on your inputs, how long it keeps data, who at the vendor can see it, and whether you can delete it. If no one can read the terms competently, get help.
4. Classify each tool. Mark each one as approved for client information, approved only for work without client information, or not approved.
5. Decide when you need client consent. For any tool that may learn from inputs, get informed consent before entering client information. Write a short, specific consent explanation. Do not rely on boilerplate.
6. Update your engagement letter. Describe how the firm uses AI in plain terms, and invite clients to give instructions.
7. Check outside counsel guidelines. Some clients prohibit or restrict AI use. Note those restrictions in the matter file.
8. Set review rules by task. Require human review of all AI output. Require citation-by-citation checking for anything filed or sent to a court, opposing counsel or a client.
9. Check court rules. Before each filing, confirm whether the court or judge requires AI disclosure or certification.
10. Label AI drafts. Marking AI-generated material in firm files tells later readers it still needs checking. This is good practice rather than a requirement of the opinion.
11. Train everyone. Cover what the tools do well, where they fail, what may not be entered, and how to report a problem. Repeat when tools change.
12. Supervise outside help. Ask contract lawyers, freelancers and vendors what AI tools they use on your matters.
13. Fix your billing practices. Bill actual time, review flat fees for tasks AI speeds up, treat general tools as overhead, and never bill for learning a routine tool.
14. Plan for incidents. Decide who staff tell if client information goes into the wrong tool, and how the firm will assess whether clients must be notified.
15. Review twice a year. Re-read vendor terms, check for new state guidance, and update the policy.

How Zrico fits

Zrico gives each firm its own private workspace, with its own application and databases. It uses Claude by default, 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 trains on your prompts, files or answers. Sign-in is invite-only with two-factor authentication, owners and admins get a usage and audit report, and retention is configurable. Zrico does not make a firm compliant with Opinion 512, and every output still needs review by a lawyer.

Sources

• ABA Formal Opinion 512, Generative Artificial Intelligence Tools (July 29, 2024): https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf
• ABA Model Rules of Professional Conduct, table of contents: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/model_rules_of_professional_conduct_table_of_contents/
• Model Rule 1.1, Competence: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_1_competence/
• Model Rule 1.4, Communications: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_4_communications/
• Model Rule 1.5, Fees: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_5_fees/
• Model Rule 1.6, Confidentiality of Information: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_6_confidentiality_of_information/
• Model Rule 3.1, Meritorious Claims and Contentions: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_3_1_meritorious_claims_contentions/
• Model Rule 3.3, Candor Toward the Tribunal: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_3_3_candor_toward_the_tribunal/
• Model Rule 5.1, Responsibilities of a Partner or Supervisory Lawyer: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_5_1_responsibilities_of_a_partner_or_supervisory_lawyer/
• Model Rule 5.3, Responsibilities Regarding Nonlawyer Assistance: https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_5_3_responsibilities_regarding_nonlawyer_assistant/
• ABA Formal Ethics Opinions, index by issue date: https://www.americanbar.org/groups/professional_responsibility/publications/ethics_opinions/aba_formal_ethics_opinions_index_by_issue_dates/

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