What the NJ Supreme Court's Preliminary AI Guidelines Actually Say About Misconduct, and What They Deliberately Leave Out

The New Jersey Supreme Court's "Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence" (January 24, 2024) is only a few pages long. If you've read it carefully, you've probably noticed something: for a document that touches on competence, client confidentiality, supervision, and candor, it spends almost no space explaining what a disciplinary consequence actually looks like when things go wrong.

That silence is worth taking seriously, especially if you're running a solo practice where there's no compliance department and no one above you to catch an error.

What the guidelines do cover

The Preliminary Guidelines identify several existing RPCs as the framework governing AI use. They point to NJ RPC 1.1 (competence), NJ RPC 1.4 (communication), NJ RPC 1.6 (confidentiality), NJ RPC 3.3 (candor toward the tribunal), and RPC 5.1 and 5.3 (supervisory responsibilities over lawyers and non-lawyer staff). These are stated as the applicable rules, not new ones created for AI.

The guidelines also set what they call an "ongoing" duty to stay current with AI developments as part of competence. The Court frames AI use as something requiring active, continuous attention, not a one-time training checkpoint.

On candor, the document is specific enough to matter: lawyers who use AI-generated content in court filings are responsible for verifying that citations and legal propositions are accurate before submission. The guidelines note that failing to do so could implicate RPC 3.3. But they stop short of specifying what a violation would look like procedurally, or whether it would be handled as a disciplinary matter versus a court-level sanction.

What the guidelines conspicuously omit

Here's the gap that practical attorneys should notice: the Preliminary Guidelines do not describe any enforcement mechanism specific to AI-related conduct. They don't establish a new reporting obligation. They don't create an AI-specific exception to existing disciplinary procedures. And they don't address what standard of proof or what level of harm would trigger a formal grievance versus a lesser administrative response.

That means the disciplinary apparatus for AI-related misconduct runs entirely through the existing Office of Attorney Ethics framework, the same process used for fee disputes, neglect, and competence failures in traditional practice. A complaint alleging that a lawyer filed AI-generated hallucinated citations without verification would go through the same intake, district ethics committee review, and potential Disciplinary Review Board process as any other RPC complaint.

The guidelines also make no mention of the Advisory Committee on Professional Ethics (ACPE). As of today, the ACPE has issued no AI-specific formal opinion. The existing opinion most relevant to data handling with online tools remains ACPE Opinion 701 (2006), which addresses electronic storage of client files and cloud-based services. That opinion requires attorneys to take reasonable precautions when using third-party electronic platforms, a standard that carries forward into AI tool use even though the opinion predates the technology by nearly two decades.

Why "preliminary" still has teeth

Some attorneys read "preliminary" as provisional, a draft that carries little weight until something more final arrives. The Court's own framing doesn't support that reading. The guidelines state explicitly that they are intended to assist lawyers in understanding how existing professional responsibility rules apply to AI. The RPCs themselves aren't preliminary. The guidelines are just the Court's current articulation of how those rules map onto a new context.

In other words, the underlying duties are fully enforceable today. An NJ attorney who files a brief containing fabricated case citations sourced from an AI tool, without checking them, faces RPC 3.3 exposure regardless of whether more specific AI rules are ever issued. The guidelines just make the Court's interpretive position clearer.

What remains genuinely unsettled is how the disciplinary system will weigh AI-related errors against context factors: Did the lawyer know the tool was prone to hallucination? Did they take any verification step at all? Did the client suffer actual harm? None of that is addressed in the Preliminary Guidelines, and no published NJ disciplinary decision has yet worked through an AI-specific fact pattern in a way that creates meaningful precedent.

What the rule requires

NJ RPC 1.1 requires competence in the law and the facts of the matter, including the tools used. NJ RPC 3.3(a)(1) prohibits making false statements of law or fact to a tribunal. The Preliminary Guidelines (January 24, 2024) affirm that both rules apply to AI-generated work product submitted to a court. No additional AI-specific rule has been adopted as of this writing.

What I recommend

Don't wait for the ACPE to issue an AI opinion before building a verification step into every AI-assisted filing. Right now, the practical minimum for a solo attorney is a written internal protocol (even one paragraph) that documents what you checked and how before filing anything an AI tool touched. If a grievance is ever filed, that paper trail demonstrates that you treated the duty seriously. Beyond that, I'd watch the Court's docket for any successor to the Preliminary Guidelines, the word "preliminary" was placed there for a reason, and a more specific set of standards is more likely to arrive with an enforcement posture attached.