Chatting With the Court: What NJ's Preliminary AI Guidelines Actually Say About Client Communication Tools

When the New Jersey Supreme Court released its Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence on January 24, 2024, most commentary fixated on the obvious: cite-checking, candor to tribunals, competence. That's understandable. But a careful read of the guidelines reveals something that solo attorneys and small firms are quietly getting wrong every day: using AI to communicate with or about clients, often without thinking through what the guidelines actually say about it.

This post is about that quieter problem.

What the guidelines actually cover

The Preliminary Guidelines are exactly what they sound like: not final rules, not ethics opinions, not disciplinary standards with independent teeth. The New Jersey Supreme Court issued them to start a conversation and give practitioners an interim framework while the court's work evolves. The ACPE has issued no AI-specific opinion as of this writing, so the Preliminary Guidelines remain the most direct official signal NJ attorneys have from the court itself.

The guidelines address several intersecting duties. They reference a lawyer's obligation to maintain competence in technology, tie that to NJ RPC 1.1, and they speak to the duty to supervise both lawyers and non-lawyer staff who use AI tools, which touches NJ RPC 5.3. But they also address something that gets far less column space: the duty to communicate with clients honestly and accurately, which sits under NJ RPC 1.4.

Specifically, the guidelines make clear that a lawyer remains responsible for the accuracy of any AI-generated content sent to a client or used on a client's behalf. That's a more consequential statement than it first appears, because a lot of AI use in small NJ firms isn't happening in briefs. It's happening in emails.

The email problem most firms don't see

Consider a hypothetical that reflects how many solos actually work: an attorney drafts a status update email for a client using an AI tool. The tool pulls context from prior case notes, generates a paragraph summarizing where things stand, and the attorney reviews it quickly before hitting send. That workflow is efficient, often accurate, and genuinely useful.

But what if the AI gets a procedural date wrong? Or frames a settlement posture in a way the client could reasonably misconstrue? Under NJ RPC 1.4(b), a lawyer must explain matters to the extent reasonably necessary to permit the client to make informed decisions. If the AI-generated explanation is subtly off, and the attorney didn't catch it because the review was cursory, the attorney still owns that error. The Preliminary Guidelines reinforce this: AI is a tool the lawyer controls, not a co-author who shares responsibility.

This matters even more when the AI tool is client-facing. Some small firms have started experimenting with AI-assisted intake questionnaires, automated matter updates, or chatbot-style FAQ pages. The guidelines don't ban any of that, but they do place the accuracy and appropriateness of everything those tools produce squarely on the lawyer. A chatbot that tells a prospective client their case has a strong chance of success based on a quick intake form? That's still the lawyer's statement, as far as the guidelines are concerned.

What RPC 1.4 asks of you here

NJ RPC 1.4(a) requires a lawyer to promptly inform the client of any decision requiring the client's informed consent and keep the client reasonably informed. RPC 1.4(b) requires explanation sufficient for informed decision-making. Neither rule distinguishes between a human-drafted communication and an AI-assisted one. The duty attaches to the communication itself, not the method of producing it.

The Preliminary Guidelines don't create a new duty here. What they do is confirm that AI doesn't create a carve-out from existing ones. A solo attorney relying on an AI tool to draft or summarize client-facing communications can't point to the tool if something goes wrong. The guidelines are explicit that the lawyer is responsible for reviewing, verifying, and taking ownership of AI-generated work.

What the rule requires

The NJ Supreme Court's Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence (January 24, 2024) state that lawyers are responsible for reviewing and verifying AI-generated content and that existing professional conduct rules apply fully to AI-assisted work. NJ RPC 1.4(a) requires prompt client notification of decisions requiring informed consent and reasonable ongoing communication. NJ RPC 1.4(b) requires explanation sufficient for the client to make informed decisions. NJ RPC 1.1 requires competence, including keeping current with relevant technology.

What I recommend

Start by auditing where AI is actually touching your client communications, not just your court filings. For most small NJ firms, the exposure isn't in briefs; it's in emails, intake summaries, and automated updates that get less review precisely because they feel low-stakes.

For any AI-generated client communication, build in a one-read verification rule: confirm dates, confirm procedural posture, confirm that the tone doesn't overstate or understate the client's situation. A 45-second check prevents the kind of miscommunication that generates a grievance.

If you're running any client-facing AI tool, such as a chatbot or automated intake form that gives substantive feedback, add a clear disclosure that the output is informational only and that the client should speak with you before acting on anything it says. That disclosure won't insulate you from a supervision failure, but it reduces the chance a client relies on AI output as legal advice.

Finally, review your AI tools' output settings. Some AI writing assistants default to confident, declarative language. That's fine for a press release. It's a problem in a client email about litigation risk. Adjust the prompts, or the template, so that uncertainty in the underlying facts shows up as uncertainty in the communication.

The Preliminary Guidelines are preliminary, but the RPCs they reference are not. Your client emails aren't a low-risk zone just because no judge reads them.