Stop Treating NJ's Preliminary AI Guidelines as a Safe Harbor, They Were Never Designed to Be One

There's a habit forming among NJ practitioners that the Court probably didn't intend to encourage. An attorney adopts an AI tool, scans the NJ Supreme Court's "Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence" (January 24, 2024), checks a few boxes in their head, and concludes they're covered. They're not wrong that the guidelines matter. They're wrong about what the guidelines actually do.

The word "preliminary" in the title isn't a formality. It signals that the Court is watching how the profession adopts AI before it decides what binding rules look like. That distinction carries practical weight for every solo or small firm using generative AI right now.

What the guidelines actually are (and aren't)

The January 2024 document is the Court's articulation of how existing RPCs apply to AI use. It doesn't create new obligations from scratch; it maps current duties onto a new technology. That includes duties of competence, supervision, confidentiality, and candor, each already embedded in New Jersey's Rules of Professional Conduct.

What the guidelines don't do: they don't give you a defined procedure that, if followed, immunizes you from discipline. There's no checklist the Court blessed. There's no certification. There's no "if you reviewed vendor terms and got written consent, you're fine" clause. The guidelines are interpretive guidance issued by the Court, not a compliance protocol issued by a regulator.

This matters because attorneys who treat them as a safe harbor may be calibrating their AI practices to the floor of what the Court described in early 2024, rather than to the ongoing duties the Court was describing. The Court made clear in the guidelines that obligations under the RPCs apply "throughout the representation", not just at onboarding or initial tool selection.

The ACPE has not spoken specifically on AI

The Advisory Committee on Professional Ethics has not issued an AI-specific opinion as of this writing. ACPE Opinion 701 (2006) addressed electronic storage and cloud-based practice management, and it remains the operative ethics opinion for questions about storing client data outside the firm. But Opinion 701 predates generative AI by nearly two decades. It wasn't written with large language models in mind, and it doesn't answer questions about what happens to client data when it's used as a prompt input.

That silence is consequential. It means there is no ACPE opinion an NJ attorney can cite to show they followed authoritative NJ-specific ethics guidance on generative AI use. The January 2024 guidelines are the only Court-issued document in that space, and they're preliminary.

What "preliminary" probably means for what comes next

Courts that issue preliminary guidance typically follow up with either formal rule amendments, a standing committee report, or binding opinions. The NJ Supreme Court has done all three in other contexts. The preliminary label here strongly suggests that the guidelines will be superseded, either by a revised version, a formal rule change to the RPCs, or both.

That creates a planning problem for small firms that haven't built any flexibility into how they use AI. If your current AI workflow was designed to satisfy the January 2024 guidelines exactly as written, and the Court tightens a specific duty, say, what "reasonable efforts" to verify AI output actually requires, you'll need to rebuild that workflow. Firms that built their practices around the underlying RPCs rather than around the specific language of the preliminary guidelines will have an easier adjustment.

The practical difference: anchor to RPC 1.1's competence duty and RPC 5.3's supervision requirements as the durable obligations, and treat the guidelines as the Court's current illustration of how those duties apply to AI. When the illustration updates, your anchor doesn't move.

What the rule requires

The NJ Supreme Court's Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence (January 24, 2024) articulate how existing NJ RPCs, including duties of competence, supervision, confidentiality, and candor, apply to attorney use of AI tools. The guidelines are issued by the Court and reflect the Court's interpretive position; they are not a separate rule, and they do not supersede or replace the RPCs themselves. ACPE Opinion 701 (2006) requires NJ attorneys using third-party electronic storage to take reasonable precautions to ensure confidentiality and to investigate the provider's practices. No ACPE opinion addresses generative AI specifically.

What I recommend

Build your AI compliance practices around the RPCs as the durable foundation, and reference the January 2024 guidelines as the Court's current application of those rules, not as a standalone checklist. Specifically: document why each AI tool you use satisfies your competence and supervision duties under the underlying RPCs. That documentation will remain relevant even after the preliminary guidelines are revised or replaced. Review ACPE Opinion 701 before storing any client information with an AI vendor, and confirm that your vendor review addresses the confidentiality inquiry the opinion requires. Finally, set a calendar alert for Q1 each year to check whether the NJ Supreme Court has issued revised guidelines or whether the ACPE has published an AI-specific opinion, because when either of those happens, any workflow you built to match the current preliminary text will need a fresh look.