Short answer: yes, with your brain still on. The New Jersey Supreme Court has not banned generative AI. It has also not given anyone a permission slip to paste a client into a chatbot and file whatever comes back.
On January 24, 2024, the Court issued Preliminary Guidelines on the Use of Artificial Intelligence by New Jersey Lawyers (Notice to the Bar, n240125a). In March 2026 it followed with a Notice on Responsible Use of Artificial Intelligence and a sample starter policy (n260330c). Both live on njcourts.gov. Read those PDFs the week you write a firm policy. This post is a map, not a substitute.
There is still no numbered Advisory Committee on Professional Ethics opinion that I can point you to as “the AI opinion.” The Court is speaking through Notices to the Bar. That is enough to take seriously. It is not a new RPC.
The RPCs did not change. The homework did.
The 2026 starter template is blunt: the firm and the supervising lawyer remain responsible for work product that incorporates AI output, and for the Rules of Professional Conduct, including RPCs 1.1 (competence), 1.6 (confidentiality), 5.1 and 5.3 (supervision of lawyers and nonlawyer assistance), and 3.3 (candor toward the tribunal). The January 2024 guidelines said the same thing in plainer clothes: existing rules apply to new tools.
The Court’s sample policy also says AI is a tool, not a decision-maker. It may help you draft or summarize. It does not replace lawyer judgment. If that sentence feels obvious, good. Write it on the sticky note anyway. The trouble starts when a tired associate treats the obvious as optional.
What the Court actually wants you to do
The March 2026 notice frames a written AI policy as a practice standard, not a magic shield. Adopting a template is not a safe harbor. The sample still asks a firm to:
- understand the capabilities and limits of the tools it chooses
- protect client and other confidential information
- review AI-assisted work for accuracy before anyone else sees it
- train and supervise the humans who click the buttons
- revisit the policy as the tools move
Human review is not a vibe. The sample says no AI-assisted content goes to a client, to opposing counsel, or to the court without lawyer review and editing. It also says you review before you rely on the output internally for legal, factual, strategic, or client-matter work. That includes the memo you only showed your partner.
Competence now has a computer in it
RPC 1.1 is still competence. You do not need to become an engineer. You do need to know, at a level that is reasonable for the matter, what the tool cannot do. The January 2024 guidelines flag a specific failure mode: generative systems can “hallucinate” and produce convincing, false information. If you cannot explain that risk to a first-year, you are not ready to let the first-year prompt it.
In March 2026 the Court also amended the CLE rules (see Notice n260401b) so that technology-related subjects, including AI, sit inside the ordinary credit requirement. Confirm the effective date and the exact hour count on the live notice. Then take the credit. A webinar is not a policy. It is how you stop pretending the printer is the only machine in the office that can embarrass you.
Where this sits next to real work
I help New Jersey firms set up private AI for confidential documents, on site, through Bonnie Law Consulting. That is a product, not a court rule. A public chatbot is not a confidential office. Grok and similar general tools can be useful for public law, outlines, and “what does this statute even mean.” They are not a sandbox for a client’s medical file.
If you only needed one sentence from the Court: you may use the tool; you may not outsource the judgment, the confidentiality, or the signature block.
Download n240125a and n260330c. Put a policy on paper. Pick one confidentiality rule and mean it. Then go outside. The hydrangeas do not hallucinate case names, which is more than I can say for a Friday-night prompt.