ABA Model Rules (AI Context)

The American Bar Association's Model Rules of Professional Conduct are advisory standards that set the baseline for attorney professional responsibility in the United States. Individual states adopt and often modify these rules, making them binding in the jurisdictions where attorneys are licensed. The ABA Model Rules were not written with artificial intelligence in mind — they were developed in 1983 and have been periodically amended — but several rules apply directly and consequentially to how attorneys use AI tools in their practice.

The most significant AI-relevant rules are:

Rule 1.1 (Competence): Requires that attorneys provide competent representation, including the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. Comment 8 to Rule 1.1, added in 2012, states that competence includes "keeping abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology." This comment is now widely interpreted to require that attorneys understand the AI tools they use — not necessarily at an engineering level, but sufficiently to use them responsibly and recognize their limitations.

Rule 1.6 (Confidentiality of Information): Prohibits attorneys from revealing information relating to the representation of a client unless the client gives informed consent, disclosure is impliedly authorized to carry out the representation, or certain enumerated exceptions apply. Rule 1.6(c) adds that attorneys must make "reasonable efforts to prevent the inadvertent or unauthorized disclosure" of client information. This rule directly governs what client data attorneys may enter into AI tools and what due diligence they must conduct about how those tools handle data.

Rule 3.3 (Candor toward the Tribunal): Prohibits attorneys from making false statements of fact or law to a tribunal and from offering evidence the attorney knows to be false. Submitting AI-generated case citations without independent verification — and having those citations turn out to be hallucinated — has been held in multiple courts to violate Rule 3.3.

Rule 5.1 (Responsibilities of Partners, Managers, and Supervisory Lawyers): Requires that firm partners and supervisory attorneys make reasonable efforts to ensure that the firm's policies and procedures result in conformity with the Rules of Professional Conduct. This rule places responsibility on firm leadership for the AI practices of subordinate attorneys — a junior associate's misuse of AI is a supervision failure by the partner if the firm has no AI governance policies.

Rule 1.5 (Fees): While not primarily an AI rule, it has AI implications: charges to clients must be reasonable. Questions have arisen about whether time that was eliminated by AI assistance should be billed at the full rate previously associated with that work, and whether AI tool costs should be passed to clients as expenses.

The ABA Model Rules create personal professional responsibility obligations for every attorney, not just organizational obligations for law firms. This means that when an attorney uses an AI tool that exposes client data — because the attorney did not investigate the tool's data handling — or files a brief containing hallucinated citations — because the attorney did not verify AI research outputs — the attorney personally faces bar discipline, court sanction, and potential malpractice liability.

The filing of AI-hallucinated case citations has already produced documented court sanctions in multiple cases. Mata v. Avianca (SDNY, 2023) was the case that brought this issue to widespread attention: the attorneys received monetary sanctions and were required to notify the judges of other cases in which they appeared. Subsequent cases in multiple federal and state courts have applied similar sanctions. The common thread is that attorneys filed AI-generated citations without checking whether the cases actually existed.

For firm management, Rule 5.1 supervision obligations mean that establishing an AI governance framework is not optional — it is a professional conduct requirement. A firm that allows its associates to use AI tools without any policy guidance has failed to make "reasonable efforts" to ensure that those associates' AI use conforms to the Rules of Professional Conduct.

ABA Formal Opinion 512, issued in July 2023, is the ABA's most comprehensive guidance on AI ethics to date. It is not binding but is highly influential — state bars routinely consider ABA formal opinions in developing their own guidance. Multiple state bars have since issued their own opinions that track or expand on Opinion 512's framework.

Note that ABA Model Rules are advisory: they must be adopted by individual states to be binding. Some states have adopted the ABA rules verbatim; others have made modifications. California operates under its own professional conduct rules that differ in some respects from the ABA Model Rules. Attorneys must follow the rules of the specific jurisdictions in which they are licensed.

How It Works

Competence under Rule 1.1 in the AI context means understanding enough about an AI tool to use it responsibly. This does not require technical expertise in machine learning — it requires knowing: what the tool is designed to do and what it is not designed to do; what kinds of errors the tool makes and at what rate; that AI outputs require verification, particularly for factual and legal claims; and that AI cannot exercise professional judgment and cannot substitute for the attorney's independent analysis. State bar guidance on AI competence is proliferating; attorneys should consult their specific bar's opinions.

Confidentiality under Rule 1.6 in the AI context requires pre-use investigation of any AI tool that will receive client information. The investigation must determine: Does the vendor train on customer data? Who are the vendor's subprocessors? What is the data residency? Is there a GDPR-compliant DPA (for EU client data)? Does the vendor commit to zero data retention after processing? The depth of investigation required is proportional to the sensitivity of the client data and the scale of use. Tools like CoCounsel, Harvey AI, and Paxton AI have specifically designed their products to address attorney confidentiality concerns and offer zero-data-retention commitments.

Candor under Rule 3.3 in the AI context requires independent verification of AI-generated factual claims, legal citations, and procedural information before those materials are filed with a court. This means: every AI-generated case citation must be verified against an authoritative primary source (Westlaw, Lexis); every AI-generated factual representation must be verified against underlying evidence; and every AI-generated procedural statement (filing deadlines, jurisdictional facts) must be verified against official rules and records. "I used an AI and it said so" is not a defense to a Rule 3.3 violation.

Supervision under Rule 5.1 in the AI context requires law firm partners and supervisory attorneys to: establish written policies governing AI use (what tools are approved, for what purposes, with what verification requirements); train subordinate attorneys and staff on these policies; and monitor compliance with the policies. The absence of AI governance policies is itself a potential Rule 5.1 violation for firms where supervisory partners knew AI was being used without oversight.

Key Considerations for Law Firms

State bar variations matter. While ABA Formal Opinion 512 provides a useful framework, attorneys must follow the rules and opinions of their specific state bars. California's rules differ from New York's, which differ from Texas's. Attorneys licensed in multiple jurisdictions must satisfy the requirements of each.

Client consent may sometimes be required. ABA Formal Opinion 512 notes that in some circumstances, attorneys may need to disclose AI use to clients and obtain informed consent — particularly when AI use materially affects the representation or when the client has expressed preferences about how their information is handled. Proactively addressing AI in engagement letters is increasingly standard practice.

The competence floor is rising. As AI tools become more prevalent and bar guidance becomes more specific, the level of AI competence that satisfies Rule 1.1 will increase. An attorney who does not know what an AI tool does with their client's data, or who does not know that AI legal research outputs require verification, may soon face competence challenges that were not viable five years ago.

Rule 1.5 fee implications require attention. Whether AI-displaced time can still be billed at previous rates, and whether AI tool subscription costs can be passed to clients as disbursements, are questions that individual firm billing policies and client engagement letters must address. There is no uniform answer, but clarity in advance avoids disputes.

Limitations and Risks

ABA Model Rules are advisory, not federal law. The ABA has no enforcement authority. Enforcement occurs through state bar disciplinary boards, which apply the rules adopted in their jurisdictions. An ABA formal opinion does not itself create a binding obligation — it is persuasive guidance that state bars may or may not follow.

Rules written before AI may not map cleanly. The Rules were developed in the context of human professional judgment and traditional legal workflows. Applying them to AI-assisted practice requires interpretive judgment, and the correct interpretation of some provisions in AI contexts is still being developed by bar regulators and courts.

Bar opinions are still catching up. While over 20 state bars have issued AI-related ethics guidance as of 2025, many have not. Attorneys in jurisdictions without specific AI guidance must apply existing rules by analogy — which involves interpretive uncertainty.