Sanctions, Supervision, and the Search History You'd Rather Not Explain: Misconduct Risk Under NJ's Preliminary AI Guidelines

Most conversations about the NJ Supreme Court's "Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence" treat the document as a checklist: verify AI output, disclose when required, supervise non-lawyers. That framing isn't wrong, but it misses the harder question. What actually happens when an NJ solo attorney follows the Guidelines imperfectly, something goes sideways in a client matter, and the Office of Attorney Ethics comes calling?

The Guidelines themselves are silent on consequences. They are framed as preliminary and educational, not as disciplinary rules. But the disciplinary rules already exist, and they don't pause for a nascent technology.

The gap the Guidelines don't fill

Issued on January 24, 2024, the Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence identify the RPCs most likely to be implicated by AI use, including RPC 1.1 (competence), RPC 1.4 (communication), RPC 1.6 (confidentiality), and RPC 3.3 (candor). What they do not do is create a new safe harbor. Following the Guidelines does not, by itself, establish that an attorney met the standard of conduct the RPCs require.

This matters because NJ RPC 8.4(c) prohibits conduct involving dishonesty, fraud, deceit, or misrepresentation, and RPC 8.4(d) prohibits conduct prejudicial to the administration of justice. Neither provision asks whether an attorney meant to deceive or whether the deception was AI-generated. The question is what happened and who was responsible.

Consider a hypothetical: a solo family law attorney uses a generative AI tool to draft a brief, doesn't verify a cited case, and that case turns out not to exist. The attorney submits the brief. Under NJ RPC 3.3(a)(1), a lawyer shall not knowingly make a false statement of fact or law to a tribunal. "Knowingly" carries a defined meaning under the Rules, but courts and disciplinary authorities have held that willful blindness, signing without reading, can substitute for actual knowledge in analogous contexts. The Preliminary Guidelines specifically instruct attorneys to verify AI-generated content before submission. Skipping that step while certifying a brief is a problem the Guidelines explicitly anticipated.

What the OAE process actually looks like

The NJ Office of Attorney Ethics receives grievances, investigates, and can refer matters to a District Ethics Committee for a hearing. If the DEC finds a violation, it forwards findings to the Disciplinary Review Board, which recommends discipline to the NJ Supreme Court. The Court issues the final order.

Discipline ranges from an admonition (private) to reprimand (public) to suspension to disbarment. For first-time, unintentional errors, admonitions or reprimands are more common. But "I didn't know the AI hallucinated" is not an independent defense, any more than "I didn't read the contract before I signed it" is a defense to a malpractice claim.

The NJ Supreme Court has made clear, in its own guidance framework, that responsibility for AI output rests with the attorney. The Preliminary Guidelines state directly that lawyers are "responsible for all work product, regardless of the tools used to create it." That language tracks the well-established principle under RPC 5.1 and 5.3 that delegation doesn't dilute accountability, but it also applies with full force when the tool is software rather than a paralegal.

The specific scenario that most concerns me

What worries me more than the hallucinated-citation scenario is the quieter one: an attorney who uses AI to generate client communications, doesn't review them carefully, and those communications contain incorrect legal information that the client relies on. Under RPC 1.4(b), a lawyer shall explain a matter to the extent reasonably necessary to permit the client to make informed decisions. If AI-drafted communications create a false picture of the case, the attorney has an RPC 1.4 problem regardless of which tool created the text.

The Preliminary Guidelines flag RPC 1.4 as one of the implicated rules but offer limited operational guidance on what adequate review of AI-generated client communications looks like. That's a gap worth filling now, before a grievance fills it for you.

What the rule requires

Under the NJ Supreme Court's Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence (January 24, 2024), NJ attorneys are responsible for all work product regardless of the tools used to produce it. The Guidelines identify RPC 1.1, 1.4, 1.6, and 3.3 as the provisions most directly implicated by AI use in legal practice. The Guidelines do not create new disciplinary standards; they describe how existing RPCs apply.

NJ RPC 8.4(c) prohibits conduct involving dishonesty, fraud, deceit, or misrepresentation. NJ RPC 8.4(d) prohibits conduct prejudicial to the administration of justice. NJ RPC 3.3(a)(1) prohibits knowingly making a false statement of fact or law to a tribunal. These rules apply to AI-assisted work product as they would to any other.

What I recommend

Build your AI review workflow around the specific failure mode you're trying to prevent, not around the Guidelines as a checklist. For court filings, that means every cited authority is opened and confirmed before the brief is signed. Full stop. For client communications, it means reading the entire output before it goes out, not skimming for tone.

Keep a brief internal log of which AI tools you used for a given matter and what review steps you took. This isn't required by any current NJ rule, but if a grievance is filed, the difference between "I had a consistent review process" and "I generally reviewed things" is the difference between a credible defense and an uncomfortable deposition.

If you're using AI for client-facing written advice in particular, treat the output the same way you'd treat a first draft from a junior associate: useful, but not ready to send without your name on it in a meaningful way.

The Preliminary Guidelines are a reasonable starting point. They are not a finish line, and the OAE's jurisdiction doesn't depend on whether the Court has finalized them.