After the ABA's 2025 AI Guidance, Here's What NJ Small Firms Actually Need to Change

In early 2025, the American Bar Association issued updated formal guidance clarifying how existing Model Rules apply when lawyers use AI tools — covering competence, supervision, confidentiality, and candor. It wasn't a bombshell. But it was a calibration point. And for NJ solo attorneys and small firms running lean, it's worth asking a blunt question: does your current AI setup actually hold up against what's now written down?

The short answer for most small NJ practices is: partially. Some things are fine. Others are quiet vulnerabilities that no one has flagged yet.

What the ABA's 2025 Guidance Actually Said

The ABA stopped short of banning any specific tool. What it did was codify a framework most ethics commentators had been inferring anyway: using AI in legal work triggers your existing duties — it doesn't create new ones, but it also doesn't suspend the old ones.

Specifically, the guidance emphasized:

• Competence (Model Rule 1.1): Lawyers must understand the AI tools they use sufficiently to recognize their outputs as reliable or unreliable. "I used AI" is not a defense to a competence failure.
• Supervision (Model Rules 5.1 and 5.3): Whether AI functions more like a non-lawyer assistant or an automated system, the supervising attorney is responsible for the work product it generates.
• Confidentiality: Inputting client data into third-party AI systems requires the same analysis as any other third-party disclosure — and consent or policy backing may be required.
• Candor: If AI-generated content is submitted to a tribunal, the lawyer is responsible for its accuracy. Period.

None of this is new in principle. What's new is that it's now in formal ABA guidance, which New Jersey courts and the NJSBA ethics committee can — and likely will — cite.

Where New Jersey Adds Its Own Layer

NJ practitioners don't just answer to the ABA. New Jersey's Rules of Professional Conduct, administered by the Supreme Court of New Jersey, track the Model Rules closely but carry independent weight. A few NJ-specific considerations sharpen the stakes:

NJ RPC 1.6 and Data Residency: New Jersey's confidentiality rule is interpreted to require reasonable precautions against unauthorized disclosure. The NJSBA's own guidance has flagged cloud-based AI tools as requiring scrutiny — not just a blanket prohibition, but a genuine vendor evaluation. If you're using a general-purpose AI tool with data stored on servers outside your control and you haven't reviewed the vendor's data use policy, you're exposed.

NJ Court-Filed Documents: Several NJ federal and state court standing orders now require or are beginning to require attorney certification regarding AI-generated content. If you're in federal district court in NJ and haven't checked your judge's standing order lately, check it this week. The posture is shifting from "courts might ask" to "courts are asking."

The NJSBA's Pending Formal Opinion: As of mid-2025, the New Jersey State Bar Association's Committee on the Unauthorized Practice of Law and the Professional Responsibility rules committee have both received requests for formal opinions on AI use. No formal opinion has been issued yet — but that absence cuts both ways. It means there's no safe harbor opinion to point to, and it means firms that establish documented, defensible practices now will be better positioned when formal guidance does arrive.

Three Practical Adjustments for NJ Small Firms Right Now

You don't need to overhaul anything. But three targeted adjustments address the highest-probability gaps:

1. Document your AI tool justification, not just your tool selection. It's not enough to say "we use [Tool X]." Your file or firm policy should reflect why that tool was selected — what you reviewed about its data handling, whether it has a SOC 2 Type II report, whether you reviewed its terms of service for training data use. This is the paper trail that answers the supervision question under both RPC 5.1 and 5.3.

2. Build a one-step verification habit into any AI-assisted filing. Before any AI-assisted document goes to a client or a court, it should pass through a specific attorney review checkpoint — not a general "look it over," but a structured verification for hallucinated citations, incorrect case holdings, and factual inaccuracies. This takes five minutes when you build it into your workflow. It takes five months to resolve if you don't.

3. Update your engagement letter to reflect AI use in the representation. The ABA guidance and the NJ framework together strongly support disclosure. Your engagement letter doesn't need to be lengthy about it — a single paragraph disclosing that your firm uses AI-assisted tools as part of its workflow, subject to attorney review and supervision, is sufficient for most matters. It handles the confidentiality analysis, documents informed consent, and signals to the client that you're practicing thoughtfully.

The Firms That Will Have Problems

The practices most at risk aren't the ones using AI aggressively. They're the ones using it casually — running client facts through a general-purpose chatbot without a data policy, relying on AI-drafted motions without a verification step, or assuming that because no one has complained yet, the workflow is defensible.

The ABA's 2025 guidance didn't change the rules. It clarified that the rules were always watching. NJ practitioners who take that seriously now — before the NJSBA issues its own formal opinion and before the first NJ disciplinary matter citing AI use becomes public — are in the best position to practice confidently with these tools.

That's the actual edge here. Not the AI. The governance around it.