ABA Formal Opinion 512: Ethics Guidance for Generative AI
The national framework. Seven Model Rules interpreted as applied to GenAI, imposing affirmative duties on every lawyer using AI in client work.
The Opinion in One Paragraph
On July 29, 2024, the ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512 — the ABA's first comprehensive ethics guidance on lawyers' use of generative AI.
It interprets seven existing Model Rules (1.1, 1.6, 1.5, 3.3, 5.1, 5.3, 7.1) as applied to GenAI, imposing affirmative duties on every lawyer using AI in client work. The opinion does not prohibit GenAI — it requires demonstrated competence, protected confidentiality, supervised use, verified output, honest billing, and truthful communication about AI capabilities.
The Seven Model Rules Op. 512 Interprets
Each rule generates an affirmative duty when GenAI is in scope
Competence
Affirmative duty to gain reasonable understanding of the GenAI tools used in legal practice — capabilities, limitations, training, data handling, hallucinations, bias, currency of training data.
Why it matters: A partner who tells associates 'use ChatGPT' without understanding the tool fails Rule 1.1.
Confidentiality
Evaluate disclosure risk before inputting client information. Informed client consent required for self-learning tools or tools with inadequate retention terms.
Why it matters: Drives the analysis in /legal-services/attorney-client-privilege-ai/.
Fees
Bills for AI-assisted work must reflect actual costs and actual labor — not the labor that would have been required without AI.
Why it matters: Cannot charge an associate's hourly rate for two hours of work AI completed in five minutes.
Candor to Tribunal
Verify every AI-generated citation before filing. False statements of fact or law to a tribunal — including hallucinated citations — violate the rule regardless of attorney intent.
Why it matters: Mata v. Avianca and its progeny are the operational reality this rule was written to capture.
Supervision
GenAI treated as analogous to a nonlawyer assistant under Rule 5.3. Partners must ensure associates using AI understand limits and verify output. Firms must train and supervise nonlawyer staff using AI.
Why it matters: Wadsworth v. Walmart (Feb. 2025) sanctioned the supervising attorney alongside the filer.
Communications About Services
Marketing language implying AI does work it does not, or implying AI work product carries human-prepared reliability, can violate the rule against false or misleading communications.
Why it matters: Firm marketing teams now need an AI capability review like the SEC's Delphia framework imposes on RIAs.
Rule 1.1 — Competence
Op. 512 reads Rule 1.1 as imposing an affirmative duty to understand the GenAI tools used in legal practice. That includes capabilities, limitations, training, data handling, and known failure modes — hallucinations, bias, and currency of training data.
A partner who tells associates "use ChatGPT for research" without understanding the tool, or an associate who runs queries without understanding RAG limits and hallucination rates, both fail Rule 1.1.
Practical Compliance — What a Firm Should Actually Do
Seven artifacts that translate Op. 512 into operational practice
Written GenAI policy
Classifies acceptable tools, prohibited inputs, supervision requirements, and verification expectations across the firm.
Engagement letter updates
Addresses GenAI use and obtains client consent appropriate to the firm's tooling.
Annual Rule 1.1 training
Every attorney using GenAI is trained on tools, limits, hallucinations, and verification protocols. Refresh annually.
Verification workflow
Citation-checking on every brief, contract, or memo that touches a tribunal or third party. Documented attorney sign-off.
Supervision documentation
Rule 5.1 and 5.3 — partners sign off on associate-prepared AI-assisted work product; nonlawyer staff supervised on AI use.
Billing policy
Rule 1.5 — bills reflect actual labor on AI-assisted work. Flat-fee or value-based billing endorsed for routine AI-assisted work.
Marketing review
Rule 7.1 — firm communications about AI capabilities reviewed for accuracy against operational reality.
ABA Formal Opinion 512 — FAQ
What is ABA Formal Opinion 512?
ABA Formal Opinion 512, "Generative Artificial Intelligence Tools," was issued July 29, 2024 by the ABA Standing Committee on Ethics and Professional Responsibility. It is the ABA's first comprehensive ethics guidance on lawyers using generative AI, interpreting seven Model Rules (1.1, 1.6, 1.5, 3.3, 5.1, 5.3, 7.1) as applied to AI.
Does ABA Op. 512 ban lawyers from using AI?
No. The opinion does not prohibit AI use. It requires lawyers using GenAI to demonstrate competence in the tools, protect client confidentiality (with informed consent where required), supervise associates and staff, verify AI output before filing, bill honestly for AI-assisted work, and communicate truthfully about AI capabilities.
Does Op. 512 require client consent before using AI on a matter?
Sometimes. Informed client consent is required before inputting client-identifying information into self-learning GenAI tools or tools with retention terms not adequate to protect confidentiality. Firms using enterprise-grade tooling with appropriate contractual terms can often satisfy the consent requirement through engagement-letter language.
How does Op. 512 treat AI under the supervision rules?
The opinion explicitly treats GenAI as analogous to a nonlawyer assistant under Rule 5.3. Supervising attorneys must have reasonable assurance the AI's work is properly directed and verified. Failures to verify subordinate work that incorporates AI output have been sanctioned under Rule 5.1 (Wadsworth v. Walmart, Feb. 2025).
Does Op. 512 affect how firms bill for AI-assisted work?
Yes. Rule 1.5 (reasonable fees) requires bills to reflect actual costs and actual labor — not the labor that would have been required without AI. Firms cannot bill an associate's hourly rate for work AI completed in minutes. The opinion endorses flat-fee or value-based billing and transparent time-tracking on AI-assisted matters.
Related Resources
Continue across the silo or bridge to a core hub
Attorney-Client Privilege and AI
How Rule 1.6 and the privilege-waiver analysis interact under Op. 512
Read article →AI Hallucinations in Legal Practice
Rule 3.3 candor and the supervision rules in live sanctioning orders
Read article →AI Contract Review Governance
Rule 5.3 in practice — supervising AI as a nonlawyer assistant
Read article →The 90-Day Governance Path
Compressed timeline to stand up an Op. 512–aligned program
Read article →Governed AI Platform Checklist
Platform features that satisfy competence, confidentiality, and audit-log duties
Read article →Operationalize ABA Opinion 512 Across Your Firm
Free Shadow AI Risk Check audits your policy, your engagement letters, your verification workflow, and your Rule 5.1 / 5.3 supervision documentation.