The American Bar Association has published its first comprehensive formal ethics opinion on artificial intelligence – and if you practice law, manage a firm, or hire lawyers, Formal Opinion 512 reshapes the calculus on AI adoption.

Not because it says anything radical. Because it says something authoritative: the Model Rules of Professional Conduct already govern how lawyers use AI tools. There is no grace period, no waiting for new regulations, no ambiguity about whether existing duties apply. They do. They always did. The ABA has now made that explicit – and in doing so, has drawn a clear line between lawyers who use AI competently and those who are exposed. While ABA Formal Opinions are persuasive rather than binding authority, they carry significant weight with state bars and disciplinary bodies, and they signal the direction of professional consensus.

This opinion arrives at a critical moment. Generative AI tools are embedded in legal workflows across the profession – from contract review and legal research to client communication and litigation strategy. Many firms adopted these tools faster than they adopted policies governing them. Opinion 512 closes that gap by mapping four existing Model Rules directly onto AI use, making clear that ignorance of how these tools work is not a defensible position.

Competence Now Includes Technological Literacy

The core of Opinion 512 is its reading of Model Rule 1.1 – the duty of competence. The ABA’s position is unambiguous: a lawyer who uses an AI tool without understanding its capabilities, limitations, and failure modes is not meeting the competence standard. You do not need a computer science degree. You do need to understand that large language models can fabricate case citations, that outputs reflect training data biases, and that no AI tool is a substitute for independent legal judgment.

This is not a hypothetical concern. Federal courts have already sanctioned attorneys for submitting AI-generated briefs containing fabricated citations – cases that did not exist, holdings that were invented. Opinion 512 makes clear that the duty to verify is not optional, and that reliance on AI output without independent confirmation is a competence failure, not a technology failure.

The practical standard the opinion establishes is analogous to expert witness oversight: you would not put an expert on the stand without understanding their methodology and vetting their conclusions. The same rigor applies to AI. If you cannot explain how the tool arrived at its output, you are not competent to rely on that output in a professional capacity.

Supervision Has a New Dimension

For partners and supervising attorneys, Opinion 512 extends supervisory duties into AI governance. Model Rule 5.1 requires supervisory lawyers to ensure that lawyers under their direction comply with the Rules; Model Rule 5.3 extends analogous duties to nonlawyer assistants. Together, they mean that if lawyers and staff at your firm are using AI tools – and they are – you have a supervisory obligation to ensure they are using them appropriately.

That obligation translates into concrete requirements. Firms need written policies governing which AI tools are approved for use and under what circumstances. They need training programs that go beyond “here’s how to log in” and address the specific risks of AI in legal work: hallucination, bias, data leakage, and overreliance. And they need monitoring mechanisms that provide accountability without creating a surveillance culture.

The firms that will be best positioned are those that treat this as a governance question, not a technology question. The tools will change. The obligation to supervise competently will not.

Confidentiality Is the Highest-Stakes Issue

Opinion 512’s treatment of Model Rule 1.6 – confidentiality – is where the practical stakes are highest. Every time a lawyer inputs client information into an AI tool, they are transmitting confidential data to a third-party system. The question is whether that transmission is consistent with the duty to protect client confidences.

The answer depends entirely on the specific tool, the specific vendor, and the specific terms governing data handling. Some AI platforms use inputs to train future models – meaning client information entered into one session could influence outputs generated for others. Some platforms retain data indefinitely. Some route data through servers in jurisdictions with different privacy protections.

Opinion 512 requires lawyers to conduct due diligence on these questions before using the tool – not after a breach, not after a bar complaint. That means reading the terms of service (actually reading them), understanding data retention and training policies, confirming that adequate technical safeguards exist, and obtaining informed consent from clients when the circumstances warrant it.

This is not a burden unique to AI. Lawyers have always been required to vet the vendors they use for document storage, e-discovery, and communication. AI tools are simply the newest category of vendor – and the one where the risks of unexamined adoption are greatest.

What This Means If You Hire Lawyers

Opinion 512 is addressed to the profession, but its implications extend to every organization that relies on legal counsel. If your outside firm is using AI tools on your matters – and increasingly, they are – you have standing to ask how. What tools are being used? What data is being input? What policies govern their use? How are outputs verified?

These are not adversarial questions. They are the same due diligence you would apply to any material aspect of how your legal work is being handled. The firms that welcome these conversations are the ones that have done the work. The firms that bristle at them may not have.

Three Things Every Firm Should Do This Quarter

First, audit your current AI usage. Not what the firm has officially approved – what people are actually using. Consumer AI tools, browser extensions, and personal accounts are likely in play across your organization. You cannot govern what you have not inventoried.

Second, publish a written AI policy. It does not need to be fifty pages. It needs to address approved tools, prohibited uses, confidentiality safeguards, supervision requirements, and verification protocols. A one-page policy that everyone reads is worth more than a comprehensive manual that no one does.

Third, invest in training that addresses AI-specific risks. Generic technology Continuing Legal Education (CLE) programs are insufficient. Your team needs to understand hallucination, prompt injection, data leakage, and the limits of AI-generated legal analysis – in terms that connect to the work they do every day.

The Competitive Reality

Formal Opinion 512 is not just a compliance event – it is a market signal. The ABA has acknowledged that AI is a permanent part of legal practice and that the lawyers who master it ethically will outperform those who either refuse to engage or engage recklessly. The opinion gives firms a framework for adopting AI with confidence, and it gives clients a basis for evaluating which firms have done the work.

The tools are here. The ethical framework is now explicit. The question is execution.

This content is provided for informational purposes only and does not constitute legal advice, nor does it create an attorney-client relationship. The information contained herein may not reflect the most current legal developments. You should consult an attorney for advice regarding your specific circumstances. This article constitutes attorney advertising under applicable bar rules. Prior results do not guarantee a similar outcome.