Stop Treating NJ's Preliminary AI Guidelines as a Checklist: Here's What They Actually Are

The New Jersey Supreme Court's "Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence," released January 24, 2024, has been cited in bar presentations, vendor marketing decks, and LinkedIn posts ever since. Most of those citations get something important wrong: they treat the guidelines as binding ethics rules, or, just as often, dismiss them as aspirational fluff with no teeth. Neither reading is accurate, and the difference matters a lot if you're a solo or small-firm attorney trying to figure out what your actual obligations are.

What these guidelines actually are

The Preliminary Guidelines are not RPCs. They are not court rules. They are not an ACPE opinion. The New Jersey Supreme Court issued them as guidance, framed explicitly as preliminary and subject to revision as AI technology and legal practice evolve. That framing is deliberate. The Court is signaling that the field is moving too fast to lock things down in a formal rule, and it wants to influence attorney conduct now without triggering a full rules-amendment process.

What that means practically: the guidelines don't create new duties on their own, but they do something nearly as significant. They interpret existing RPCs, including NJ RPC 1.1 (competence), NJ RPC 1.4 (communication), NJ RPC 1.6 (confidentiality), NJ RPC 5.1 and 5.3 (supervisory responsibilities), and others, in the context of AI use. When the Court tells you in writing that competence under RPC 1.1 includes understanding the capabilities and limitations of AI tools you use, that interpretive statement will matter if your conduct is ever scrutinized in a disciplinary proceeding. A prosecutor or grievance panel citing the guidelines against you would have a credible argument even if the guidelines themselves aren't technically binding.

Think of it less like a rule and more like a memo from the Court explaining how it reads the rules it already has authority to enforce.

The ACPE gap and why it matters

Here's a detail that almost nobody talks about: as of this writing, New Jersey's Advisory Committee on Professional Ethics has issued no AI-specific opinion. The ACPE is the body that issues formal ethics opinions interpreting the RPCs for New Jersey attorneys. ACPE Opinion 701 (2006), the committee's opinion on electronic storage and cloud-based client data, remains the closest analogue, and attorneys reasonably use it as a framework for thinking about data security obligations when evaluating AI platforms.

But Opinion 701 was written when "the cloud" meant a third-party server hosting your document management system. It predates generative AI by more than fifteen years. Applying it to a large language model that processes your client's facts in real time requires a significant inferential leap. That gap is not a loophole. It's a signal that the existing framework was not designed for what you're doing today, and the absence of a targeted ACPE opinion means there's no authoritative NJ-specific interpretation of how RPC 1.6 or RPC 1.1 applies to, say, using ChatGPT to draft a demand letter with client details in the prompt.

The Preliminary Guidelines are, in part, filling that interpretive vacuum. That's not nothing.

The revision cycle you should be tracking

The word "preliminary" in the title isn't decorative. The Court has indicated these guidelines will be updated. If you set a calendar reminder when the guidelines came out in January 2024 and then forgot about them, you may already be operating on an outdated understanding. Any revision could tighten expectations around disclosure to clients, around what supervision of AI-generated work looks like, or around how attorneys document their AI use. Following the NJ Courts website where these guidelines were published is the simplest way to stay current, and it takes thirty seconds to check every few months.

Solo attorneys are especially exposed here. A larger firm has a general counsel or a professional responsibility partner who tracks this. If you're a solo, you are that person, and the Court's guidelines are addressed directly to you.

One thing the guidelines got genuinely right

The Preliminary Guidelines are explicit that attorneys remain responsible for AI-generated work product. That's not a new idea in general, but the Court's articulation is worth reading carefully. The guidelines connect that responsibility back to the existing supervisory framework in RPC 5.3, which governs non-lawyer assistance, and suggest that AI tools occupy a similar conceptual space. That framing has real consequences for how you review output, how you document that review, and how you'd explain your process if a client challenged the quality of your work.

Getting that framing right, and treating AI output as something that requires the same scrutiny as work from a capable but fallible paralegal, is probably the most durable takeaway from the guidelines regardless of how they're eventually revised.

What the rule requires

NJ RPC 1.1 requires competent representation, including the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. NJ RPC 1.6 prohibits revealing information relating to the representation of a client without informed consent. NJ RPC 5.3(b) requires a lawyer with direct supervisory authority over a non-lawyer to make reasonable efforts to ensure the non-lawyer's conduct is compatible with the lawyer's professional obligations. The NJ Supreme Court's Preliminary Guidelines on New Jersey Lawyers' Use of Artificial Intelligence (January 24, 2024) interpret these and other RPCs in the AI context and indicate that attorney responsibility for AI-generated work product continues to apply. ACPE Opinion 701 (2006) addresses confidentiality obligations in the context of electronic storage and cloud-based services and remains the closest NJ-specific ethics opinion applicable to AI data handling. No AI-specific ACPE opinion has been issued.

What I recommend

Read the Preliminary Guidelines as a document, not as a summary someone else gave you. They're a few pages and the Court's own language is often more nuanced than the shorthand versions circulating in bar presentations. Bookmark the NJ Courts page where the guidelines live and check it quarterly. If you're unsure how Opinion 701's framework applies to a specific AI tool you're evaluating, treat that uncertainty as a reason to ask the question before you start using the tool, not after. And if you're relying on an AI platform for anything that touches client information, document your review process in a short internal note each time. That note is your evidence that you treated the AI's output as requiring professional judgment, not as a finished product.