When Lawyers Lean on AI, Old Ethics Rules Still Apply

Imagine hiring a contractor to renovate your home, only to learn later that they used a new power tool they'd never trained on, didn't tell you about it, and charged you full price even though the tool cut the job time in half. You'd feel misled. That's roughly the situation unfolding in law offices across the country as attorneys adopt artificial intelligence tools without always pausing to consider what they owe the people they represent.

AI has become a fixture in legal practice. Attorneys now use it to research case law, draft contracts and briefs, predict how judges might rule, and sift through mountains of documents during corporate due diligence or e-discovery. The appeal is obvious: these tools promise speed and lower costs. But speed without guardrails can create real harm, and the legal profession's existing ethics framework, built long before generative AI existed, is now being asked to do double duty.

The American Bar Association's Model Rules of Professional Conduct don't set a single national standard. States adapt them locally, which means an attorney's obligations can shift depending on where they practice. Still, several core duties apply almost everywhere, and they're worth understanding even if you're not a lawyer, because they shape how AI touches your case, your money, and your private information.

Start with competence. Under Model Rule 1.1, attorneys must handle their clients' matters capably. In 2012, the ABA added language making clear that competence includes understanding the benefits and risks of relevant technology. Think of it like a doctor's duty to know about a new medication's side effects before prescribing it. An attorney who doesn't understand how an AI tool works, or where it might fail, can't responsibly rely on it.

This isn't a brand-new idea. Back in 2004, the New York State Bar Association's ethics committee said attorneys have a duty to understand technology well enough to avoid, for example, accidentally leaking client secrets hidden in email metadata. California went further in 2015, telling lawyers who lack the technical know-how for e-discovery to either learn the skill, bring in a consultant, or turn down the case entirely. The same logic now applies to AI.

States are starting to codify this expectation through continuing legal education requirements. Florida became the first state to mandate technology-focused CLE courses in 2016. North Carolina followed in 2019. New York now requires attorneys to complete at least one CLE credit hour on cybersecurity, privacy, and data protection every two years, a rule that took effect in 2023. California started requiring one hour of technology-focused training beginning with the compliance period ending January 31, 2025. New Jersey will join them starting January 1, 2027. Given how many attorneys now work remotely, expect more states to follow this path.

Client Trust and the Price of a Shortcut

Competence is only part of the picture. Clients also have a right to know what's happening with their own case, and that's where Model Rule 1.4 comes in. It requires attorneys to consult with clients about how they intend to pursue the client's goals. If an attorney plans to use AI, best practice is to say so, ideally in the engagement letter, and to explain why the tool is being used, what information will be fed into it, and how that information might be used or disclosed afterward.

This isn't a box-checking exercise. The ABA's Formal Opinion 512, issued in 2024, makes clear that clients need enough information to meaningfully participate in decisions about their own representation. A client who doesn't know an AI tool is reviewing their sensitive documents can't make an informed choice about whether they're comfortable with that. Obtaining written, informed consent before using AI is the safer path, both ethically and practically.

Then there's the question of money. Model Rule 1.5 bars attorneys from charging unreasonable fees, and reasonableness depends on factors like the time, skill, and difficulty involved in a task. Attorneys can reasonably pass along the actual cost of using an AI tool or hiring an AI-powered vendor. What they can't do is bill a client at a full hourly rate for research or document review that AI could have completed faster and cheaper. That's not a minor technicality. It's the difference between fair billing and what looks a lot like padding an invoice using outdated methods when a faster, cheaper one was available.

Confidentiality rounds out the picture, and it may be the highest-stakes issue of all. Model Rule 1.6 prohibits attorneys from disclosing client confidences without consent or legal justification. When an attorney feeds client data into an AI system, they need to think hard about who can access that data, how secure the platform really is, and what happens if the system is breached. The New York bar's 2014 opinion on cloud storage set an early marker here: attorneys must ensure only authorized people can reach systems holding confidential information, and they need to actively monitor for breaches rather than assume a vendor has it handled.

An emergency might justify taking on some risk that wouldn't otherwise be acceptable. But that's the exception, not the rule. Clients can also simply instruct their attorney not to use certain technology, and that instruction has to be respected.

Why It Matters

None of this is about slowing down innovation for its own sake. AI genuinely can make legal work faster, cheaper, and in some cases more accurate. But the profession's ethical rules exist precisely because speed and cost savings mean little if they come at the expense of a client's trust, their money, or their private information. Attorneys who treat competence, communication, fair billing, and confidentiality as optional extras rather than foundational duties risk real harm to the very people they're supposed to protect. Getting this balance right isn't just good lawyering. It's the difference between AI serving justice and quietly undermining it.