New Jersey AI Ethics Rules for Lawyers

The official instruments

Court-issued, and deliberately restrained

New Jersey’s AI authority did not come from a bar ethics committee. It came from the New Jersey Supreme Court, which issued the Preliminary Guidelines on New Jersey Lawyers’ Use of Artificial Intelligence on January 24, 2024, through a court notice. That origin is the first thing to get right about New Jersey, and the easiest to mislabel. These are guidelines, preliminary by their own terms, and this tracker treats them as guidance rather than as a numbered formal opinion. But the source is the state’s highest court, not a voluntary or even a mandatory bar’s ethics body, which gives them an institutional standing particular to New Jersey: when the court that promulgates the Rules of Professional Conduct tells lawyers how those rules apply to AI, the interpretation carries the authority of the rule-maker.

The guidelines’ defining move is restraint. Their central holding is that AI changes no duties under the Rules of Professional Conduct. New Jersey did not write AI-specific obligations; it declared that the existing duties already reach AI and then mapped them onto it. That choice is a substantive position, not an absence of one. It means a New Jersey lawyer’s AI conduct is measured against competence, confidentiality, candor, and supervision as those rules already exist, which keeps the analysis durable as the tools change underneath it. It also marks New Jersey as an early entry in the field: a January 2024 court notice put the state among the first jurisdictions to address lawyers and generative AI at all, and the tracker labels it guidance, court-issued, with a verified day-level date, exactly as the notice supports.

What the guidelines actually require

Within that no-new-duties frame, four obligations are explicit, and they map cleanly onto a workflow. Verification first: lawyers must verify all AI-generated information, the duty that the fabricated-citation problem made universal and that New Jersey states without qualification. Confidentiality next, stated as a precondition: a lawyer must ensure a tool’s security before entering non-public client information, which in practice means assessing the tool’s data handling before any client information goes near it. Candor to the tribunal third, and this is where New Jersey’s honesty about disclosure matters: candor applies in full, there is no excuse for false content reaching a court, but there is no duty to disclose AI use to a tribunal. Using AI is not itself a disclosable fact under the guidelines; submitting false material is the violation, AI or not. And supervision fourth: firm-wide AI use falls under the oversight duties of Rules 5.1 to 5.3, making the supervising lawyer responsible for how the firm and its nonlawyers use these tools.

Client communication gets its own calibrated rule, and New Jersey resists the blanket approach. There is no general duty to tell clients about AI use. Disclosure to the client is required in two situations: if the client asks, or if the client cannot make an informed decision about the representation without knowing. That is a trigger-based duty, narrower than a categorical disclosure requirement, and it places New Jersey toward the middle of the national consent spectrum rather than at either edge. For a New Jersey firm, the operational consequence is a short, concrete artifact set: a tool-security check before client information is entered, a verification step on every AI output, engagement-letter language ready for the situations that trigger client disclosure, and supervision documented under 5.1 to 5.3. The policy template implements each of these, and the disclosure comparison table shows where New Jersey’s trigger-based client rule and its no-tribunal-disclosure position sit relative to the rest of the country.

The court layer New Jersey lawyers still owe

Because New Jersey’s instrument is itself court-issued, it is tempting to assume it settles the court-facing question. It does not, and the distinction is important. The Preliminary Guidelines are the Supreme Court speaking to lawyers’ professional duties; they expressly do not require disclosing AI use to a tribunal. But individual judges can still maintain their own standing orders on AI in filings, and the files verify no separate statewide New Jersey court rule that imposes a uniform AI-in-filings requirement on top of the guidelines. So the practical court-facing obligation is judge-by-judge, particularly in the federal districts, where standing orders vary. Before any filing, confirm the assigned judge’s individual practices rather than assuming the guidelines’ no-disclosure position controls in that courtroom; the court orders guide covers that per-filing check, and the Ropes & Gray tracker is a useful cross-reference. The state-bar layer and the court layer move at different speeds, which is exactly the kind of change the changelog and alerts are built to catch.