After ABA Formal Opinion 512, What Does "Competent" AI Use Actually Look Like for a NJ Attorney?

When the ABA Standing Committee on Ethics and Professional Responsibility published Formal Opinion 512 in July 2024, it was the first time the organized bar had worked through what generative AI means for core attorney obligations in a sustained, rule-by-rule way. For NJ solo and small-firm attorneys, the opinion is worth reading carefully, not because it changes your rules, but because it maps existing professional conduct obligations onto AI use with more precision than anything the ACPE has issued to date.

A quick orientation: ABA formal opinions don't bind New Jersey attorneys. The state's Rules of Professional Conduct, adopted and enforced by the Supreme Court of New Jersey, govern NJ practice. But the ABA's analysis is highly persuasive, the ACPE regularly looks to it when drafting opinions, and courts and grievance panels cite it. Knowing what Opinion 512 says, and where NJ's rules track it, puts you in a stronger position.

What Opinion 512 actually covers

The opinion is organized around five duties: competence, communication, confidentiality, supervision, and fees. A few points are worth highlighting because they're more specific than lawyers tend to expect from bar guidance.

On competence, Opinion 512 says that under Model Rule 1.1 (which NJ mirrors in RPC 1.1), a lawyer must understand the benefits and risks of AI tools well enough to use them effectively. That's not a general technology-literacy platitude. The opinion specifies that this includes understanding how a particular AI tool generates output, what data it was trained on, whether it fabricates citations, and how to verify results. The committee used the phrase "output verification" and tied it directly to competent representation, meaning a lawyer who submits AI-generated work product without independent review has a competence problem, not just a prudence problem.

On confidentiality, Opinion 512 addresses Model Rule 1.6 directly and says a lawyer must evaluate an AI vendor's terms of service before sending any client information. Specifically, the committee focused on whether the tool uses inputs to train its model, whether data is retained and for how long, and whether the vendor's security practices are adequate. NJ RPC 1.6(a) imposes the same confidentiality obligation and has been interpreted to cover third-party service providers under the reasonable-measures standard.

On fees, Opinion 512 is clear that lawyers cannot charge clients for AI costs in ways that are unreasonable or constitute double-billing. If a task that used to take four hours now takes forty-five minutes with AI assistance, billing the client for four hours isn't permitted. The opinion also addresses the question of passing through AI subscription costs and requires transparency if those costs are being charged to the client.

On supervision, Opinion 512 addresses the duties owed when non-lawyer staff use AI tools, tying the analysis to Model Rules 5.1 and 5.3. Even where the AI tool itself performs a task, a lawyer must review the output as if a junior associate had produced it.

Where NJ attorneys should pay particular attention

NJ RPC 1.1 and its Comment [8] on technology competence closely parallel the ABA model. But NJ RPC 1.6 is in some respects stricter than the ABA model when it comes to third-party disclosure, and the ACPE has not yet issued a formal opinion squarely interpreting RPC 1.6 in the context of AI vendor data practices. That means NJ attorneys are currently relying on general principles from ACPE Opinion 701, which addressed cloud computing, and applying them by analogy to generative AI tools.

That's a real gap. Opinion 701 was issued in 2006 and references technology that looks nothing like a modern LLM. The takeaway: until the ACPE issues AI-specific guidance, Opinion 512 is the most detailed framework available, and a NJ attorney who follows its methodology for evaluating AI vendors will have a defensible record of reasonable precautions under RPC 1.6.

What Opinion 512 doesn't resolve

The opinion is candid about what it leaves open. It does not set a specific verification standard (how many sources must you check? must you run every citation through Westlaw?). It doesn't address AI use in specific practice areas where other rules apply, like advertising, client solicitation, or courtroom conduct. And it doesn't address what happens when an AI vendor unilaterally changes its data terms mid-subscription, which is a real and common problem.

For NJ small firms, the most practical read of Opinion 512 is this: the bar has now said clearly that using AI without understanding how it works is a competence problem, not just a risk-management inconvenience. The opinion is instructive, not aspirational.

What the rule requires

NJ RPC 1.1 requires a lawyer to provide competent representation, including the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. Comment [8] to NJ RPC 1.1 (as adopted by the Supreme Court of New Jersey) states that competence includes keeping current with changes in the law and its practice, including the benefits and risks associated with relevant technology. NJ RPC 1.6(a) requires a lawyer to maintain the confidentiality of information relating to the representation and to use reasonable care to prevent the unauthorized disclosure of client information by third parties, including service providers. ACPE Opinion 701 applies this standard to cloud-based service providers and requires attorneys to evaluate the provider's terms of service, security practices, and data handling before transmitting client information.

What I recommend

Start with Opinion 512's five-duty checklist as an internal audit. For each AI tool your firm uses, write down: (1) what client data, if any, goes into it; (2) what the vendor's retention and training policy says; (3) how you verify the output; and (4) how you handle billing for AI-assisted time. That four-point record doesn't have to be long, but it should exist. If you're ever in front of a grievance panel, showing that you thought through these questions in advance is far more protective than explaining that you trusted the product's marketing page. Until the ACPE issues dedicated AI guidance, Opinion 512's reasoning is the best available benchmark, and treating it as persuasive authority rather than optional reading is the right call.