Regulatory Record Overview
| Field | Value |
|---|---|
| Issuer | American Bar Association (ABA) Standing Committee on Ethics and Professional Responsibility |
| Jurisdiction | US — model guidance; adopted by state bars individually |
| Primary Rules | Model Rule 1.1 (Competence), 1.6 (Confidentiality), 5.1 (Supervisory Responsibility — Lawyers), 5.3 (Supervisory Responsibility — Nonlawyers) |
| Formal Ethics Opinion | ABA Formal Opinion 512 (July 2024) |
| Obligation Type | Guidance — competence, supervision, confidentiality |
| Primary Source | ABA Formal Opinion 512, americanbar.org |
What the ABA Has Actually Said
ABA Formal Opinion 512, issued in July 2024, is the primary document addressing generative AI use by lawyers under the Model Rules. It does not create new rules — it applies existing ones to AI-assisted practice. The opinion focuses on four rules: 1.1 (Competence), 1.6 (Confidentiality), 5.1 (Supervision of lawyers), and 5.3 (Supervision of nonlawyer assistants).
The ABA's position is that AI tools are not categorically different from other legal technology — but their capacity to produce plausible-sounding errors at scale makes the competence and supervision obligations more demanding in practice, not less.
Rule 1.1: Competence
Rule 1.1 requires lawyers to provide competent representation, which includes the "legal knowledge, skill, thoroughness, and preparation reasonably necessary." The 2012 amendment added Comment 8, specifying that competence includes keeping abreast of changes in the law and "the benefits and risks associated with relevant technology."
Formal Opinion 512 interprets this to mean attorneys must understand, at a functional level, how the AI tools they use work — not necessarily the underlying model architecture, but enough to recognize the tool's limitations and failure modes. Submitting AI-generated legal research or drafts without independent verification is treated as a competence failure, not a mere oversight.
Rules 5.1 and 5.3: Supervision
The supervision rules are where AI creates the most novel obligations. Rule 5.3 has historically governed nonlawyer assistants — paralegals, contract reviewers, document processors. Formal Opinion 512 treats AI systems as falling within this category: a lawyer who uses an AI tool to draft, research, or analyze has supervisory responsibility for the output.
This means the same standard that applies to reviewing a paralegal's work applies to reviewing AI output. Ratifying an AI-generated product without review is the equivalent of signing off on a junior associate's brief without reading it — a clear supervision failure.
Rule 5.1 adds a layer for supervising partners and law firm management. Firms that deploy AI tools across practice groups have an obligation to establish policies and procedures that make compliance with the professional rules reasonably likely. That includes training, output verification protocols, and defined review workflows.
Rule 1.6: Confidentiality
Formal Opinion 512 treats client data input into AI systems as subject to Rule 1.6's confidentiality protections. Attorneys must evaluate whether a tool's data handling — including whether the vendor uses input data for model training, how long data is retained, and where it is stored — is consistent with their confidentiality obligations.
The opinion does not prohibit AI use, but it does require informed consent analysis: if there is a material risk that client information will be disclosed to third parties through the AI system's operation, the lawyer must either obtain informed consent or avoid using that tool for that matter.
How the Rules Map to AI Use Scenarios
| AI Use Scenario | Primary Rule | Obligation | Failure Mode |
|---|---|---|---|
| Legal research via AI (case law retrieval) | Rule 1.1 | Verify all citations against primary sources before filing | Hallucinated citations submitted to court |
| AI-drafted motion or brief | Rules 1.1, 5.3 | Attorney must review and take responsibility for substance | Fabricated arguments or misstatements of law |
| Contract review using AI analysis tool | Rules 1.1, 5.3 | Understand tool's training scope; verify flagged clauses | Tool trained on different jurisdiction's standards |
| Uploading client documents to third-party AI tool | Rule 1.6 | Evaluate vendor data retention and training practices | Client data used for model training without consent |
| Firm-wide AI deployment by managing partner | Rule 5.1 | Establish policies, training, and review procedures | No supervision framework; associates rely on AI output unchecked |
| Using AI for eDiscovery document review | Rules 1.1, 5.3 | Validate sampling methodology; human review of edge cases | Privileged documents produced due to classification error |
State Bar Activity: Where Formal Opinions Have Been Issued
The ABA's Model Rules are not binding law — they become enforceable when adopted by individual state bars, which may modify them. As of mid-2026, a significant number of state bars have issued their own AI ethics opinions, and several have diverged from the ABA's framing in notable ways.
- California State Bar: Issued Practical Guidance on AI (November 2023), followed by a more detailed formal opinion in 2024. California's guidance explicitly addresses fee disclosure when AI reduces time on a matter, going further than ABA Opinion 512 on that point.
- Florida Bar: Advisory Opinion 24-1 (2024) addresses generative AI use in legal practice, emphasizing that attorneys remain responsible for all work product regardless of how it was generated.
- New York State Bar Association: Released a comprehensive report on AI and the legal profession (April 2024) with detailed recommendations on competence, supervision, and billing — though as a bar association report rather than a formal ethics opinion, it carries advisory rather than disciplinary weight.
- Texas: The State Bar of Texas Professional Ethics Committee issued Opinion 705 (2024) addressing AI use, with particular attention to confidentiality obligations when using cloud-based AI services.
Court-Level Requirements: AI Disclosure Orders
Separate from bar ethics opinions, a growing number of federal courts have issued standing orders or local rules requiring disclosure of AI use in filed documents. These are court-specific requirements, not ethics rules — but they interact with the competence and supervision framework.
Several district courts now require attorneys to certify that any AI-generated content in filed documents has been reviewed and verified by a licensed attorney. This aligns with the Rule 5.3 supervision obligation but imposes it as a procedural requirement with contempt exposure, not just professional discipline risk.
What the ABA Has Not Resolved
Formal Opinion 512 leaves several questions open. The opinion does not specify what level of AI tool understanding satisfies the competence requirement — "functional" understanding is the standard, but its application to specific tools and use cases is left to case-by-case analysis.
- Billing: The opinion does not address whether AI-assisted work can be billed at the same hourly rate as fully attorney-performed work, or whether efficiency gains must be passed to clients. California's guidance goes further here; the ABA has not.
- Disclosure to clients: The opinion does not require disclosure to clients that AI was used, as long as the work product meets the competence standard. Some state bars have taken a different position on this.
- Agentic AI: The opinion was drafted primarily with generative AI assistants in mind. Autonomous AI agents that take actions on behalf of a client — filing, communicating, executing transactions — raise supervision questions the 2024 opinion does not squarely address.
- Third-party AI providers as "nonlawyer assistants": The Rule 5.3 analogy to AI tools is the opinion's most significant interpretive move, but its limits are untested. Courts and disciplinary bodies have not yet confronted the question of whether a vendor's AI failure can be attributed to attorney supervision failure.
Primary Source Reference
A lawyer who uses generative AI must take steps to ensure the accuracy and completeness of the AI-generated work product. Lawyers must also be mindful of their obligations of confidentiality when inputting client information into AI tools.
ABA Formal Opinion 512 (July 2024), Summary
The full text of ABA Formal Opinion 512 is available through the ABA's ethics resources portal. State bar opinions referenced above are available through each state bar's published ethics opinions database.
Comments
Join the discussion with an anonymous comment.