Supervision, Sanctions, and the Silence Between: What NJ's Preliminary AI Guidelines Don't Say About Malpractice Risk

The NJ Supreme Court's "Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence" (January 24, 2024) get cited constantly in ethics conversations, but there's a dimension of risk they barely touch: civil malpractice. That's not a flaw in the document, exactly. The Supreme Court's guidelines are an ethics instrument. They speak to disciplinary exposure. Malpractice is a different track, and the two can diverge in ways that catch solo practitioners badly off guard.

Understanding where those tracks split is, right now, one of the more practical things a small-firm attorney in New Jersey can do.

The guidelines speak to professional conduct, not to tort duty

The January 2024 guidelines remind attorneys that existing RPCs apply to AI use. They specifically call out the duty of competence under NJ RPC 1.1, which requires the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. They also invoke RPC 1.6 on confidentiality and RPC 5.3 on supervision of non-lawyer assistance.

What the guidelines do not do is define a standard of care for negligence purposes. That's a distinction with real consequences. A disciplinary tribunal asks whether you violated a rule of professional conduct. A civil court, in a malpractice case, asks whether you acted as a reasonably competent attorney would have under the circumstances. The two standards overlap significantly, but they are not the same question, and they are not decided by the same body.

So an attorney who uses an AI tool to draft a motion, reviews it cursorily, and files something that contains a material legal error may or may not face a disciplinary proceeding. She almost certainly faces a malpractice claim if the client suffers a concrete harm. The guidelines don't map that exposure for you. They were never designed to.

Why this matters for solo practitioners specifically

Solo attorneys don't have a partner looking over the AI output. They don't have a firm-wide vetting protocol developed by a general counsel's office. The review step is entirely theirs, and if it's inadequate, there's no institutional backstop.

Here's a concrete hypothetical: a solo family law attorney in Bergen County uses an AI tool to identify relevant case law on equitable distribution. The tool returns a summary that mischaracterizes the holding of a decision she doesn't independently verify. She cites the case in a certification. The court is unimpressed, the client gets a worse outcome than the facts warranted, and the client's next call is to a malpractice plaintiff's attorney.

The Preliminary Guidelines would say: RPC 1.1 required competence; the attorney had a duty to review AI-generated content. True. But the malpractice inquiry runs further. It asks what a reasonably competent family law attorney in New Jersey would have done, given the specific task, the specific tool, and the specific stakes. That inquiry will be shaped by expert testimony, not by the NJ Supreme Court's guidance document.

The ACPE has not issued an AI-specific opinion

As of the date of this post, the Advisory Committee on Professional Ethics has issued no opinion specifically addressing AI use. ACPE Opinion 701 (2006) addresses electronic storage and cloud-based practice management, and its requirements around reasonable care in vendor selection carry some analogical relevance. But there is no ACPE opinion that establishes an ethics-layer standard for AI-generated work product, which means the disciplinary track itself still has significant interpretive gaps.

That gap cuts both ways. An attorney isn't on clear notice of an established rule being violated when the rule's application to the specific technology hasn't been spelled out. But a malpractice plaintiff's expert doesn't need an ethics opinion to testify about what a competent attorney should have done.

The review obligation is where the two tracks finally meet

Both ethics exposure and malpractice exposure converge at one point: the adequacy of the attorney's review of AI output before it goes to a client, a court, or a counterparty. The Preliminary Guidelines describe this as an "ongoing" duty. Civil malpractice doctrine would likely frame it as a component of the standard of care.

That means the practical question isn't just "did I read the AI output?" It's "did I read it with the same critical attention I'd give to a first-year associate's draft?" If the honest answer is no, both tracks have a problem.

What the rule requires

NJ RPC 1.1 requires the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. The NJ Supreme Court's Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence (January 24, 2024) confirm that this duty applies to AI-generated work product and describe the obligation of review as ongoing. ACPE Opinion 701 (2006) requires reasonable care in selecting and overseeing electronic service providers handling client information, but it addresses cloud storage and does not speak directly to AI-generated work product.

What I recommend

Keep malpractice and ethics exposure on two separate mental ledgers, because they're answered by different bodies under different standards. For each AI-assisted work product that leaves your office, document what you checked and why you were satisfied with it. A three-sentence file note ("Reviewed AI draft against [source]; verified citations independently; revised paragraph three") takes ninety seconds and creates a record that matters in both disciplinary and civil contexts. If you're using AI for a task where the stakes of an error are high, such as a litigation certification or a transactional representation with financial consequences, elevate your review intensity to match. The Preliminary Guidelines won't protect you from a malpractice claim. Only your actual review will.