After the ABA's Formal Opinion 512, Here's What NJ Solo Attorneys Actually Need to Do Differently

In July 2024, the ABA released Formal Opinion 512 — its first comprehensive ethics guidance on generative AI tools in legal practice. It addressed competence, confidentiality, communication, supervision, and fees all in one document. It is, frankly, the most substantive thing any national bar body has published on this subject.

And most NJ solo attorneys have either skimmed it, bookmarked it, or ignored it entirely.

That's a problem worth correcting — because while New Jersey's own Rules of Professional Conduct don't yet have AI-specific amendments, the ABA's formal opinions carry persuasive weight in disciplinary proceedings, and the New Jersey Supreme Court's Committee on Attorney Advertising and the Office of Attorney Ethics have both signaled heightened attention to technology-related conduct. Formal Opinion 512 is the closest thing to a national standard that exists right now, and it maps directly onto New Jersey's existing RPCs.

Here's what it actually says, and what it means for your practice.

---

Competence Isn't Just About Using AI — It's About Understanding It

Opinion 512 makes clear that RPC 1.1 (competence) requires lawyers to understand not just how to use a generative AI tool, but how it functions well enough to spot its failures. That's a higher bar than most attorneys appreciate.

This doesn't mean you need a computer science degree. It means you need to understand, at minimum, that large language models hallucinate — they generate plausible-sounding but factually incorrect content, including fake citations, misquoted statutes, and fabricated case holdings. If you're using AI to assist with legal research or brief drafting and you can't describe your own verification process, you probably don't meet the competence standard.

What to do: Build a written hallucination-check step into every AI-assisted workflow. For legal research, that means independently verifying every cited case through Westlaw, Lexis, or Fastcase before it enters a document. Don't delegate this verification to a paralegal who also doesn't understand the failure modes.

---

Confidentiality Requires More Than Clicking "I Agree"

The Opinion addresses RPC 1.6 (confidentiality) by requiring attorneys to understand what AI vendors do with client data before using any tool that touches client information. This goes beyond reading a privacy policy — it means affirmatively confirming whether the tool trains on your inputs, where data is stored, whether a Business Associate Agreement is available (if health information is involved), and whether the vendor's data retention practices are compatible with your obligations.

New Jersey attorneys handling matters in regulated industries — healthcare, financial services, family law with sensitive disclosures — face heightened exposure here. NJ courts have not softened the confidentiality obligation for smaller firms.

What to do: For any AI tool you use with real client data, locate the vendor's data processing addendum (DPA). Confirm the vendor does not use your inputs for model training by default, or that you've opted out. If you cannot locate this document, treat the tool as unsuitable for client-specific work until you can.

---

Supervision of AI Output Is a Lawyer Responsibility, Full Stop

One of Opinion 512's sharpest points: AI is not a supervised person. It is a tool. The output it generates does not carry the judgment or responsibility of another professional — it carries yours. The Opinion explicitly warns against using AI-generated work product without meaningful lawyer review, which maps cleanly onto NJ RPC 3.3 (candor toward the tribunal) when AI-drafted content is filed with a court.

You've already seen what happens when this step is skipped. Federal and state courts across the country — including judges in New Jersey's federal district — have sanctioned attorneys for filing AI-generated briefs with fabricated citations. The Opinion underscores that this is not just a careless mistake; it may constitute a violation of multiple RPCs simultaneously.

What to do: Establish a rule in your practice — nothing AI-generated is filed, sent to a client, or submitted to any tribunal without an attorney reviewing the complete document for accuracy, not just grammar or tone. This is not optional and it is not delegable to staff.

---

Fees and AI Efficiency Are in Tension — And the Opinion Acknowledges It

Opinion 512 touches on RPC 1.5 (reasonable fees) in a way that should prompt billing policy reviews at every firm using AI. If AI dramatically reduces the time required to complete a task, billing the same hourly rate for the same number of hours may not hold up as "reasonable" — particularly if a client later challenges the bill.

New Jersey attorneys should be ahead of this curve. A flat-fee or value-based billing structure is increasingly defensible precisely because it decouples the fee from the time input. If you're still billing purely by the hour on AI-assisted matters without any client disclosure, you're operating in a gray zone that's getting grayer.

What to do: Review your engagement letters. Consider adding a short disclosure noting that your firm uses AI tools to improve efficiency, and that fees reflect the value of legal services rendered rather than time input alone.

---

The Bottom Line

Formal Opinion 512 doesn't create new rules. It clarifies that the existing rules — competence, confidentiality, candor, fees, supervision — already apply to everything you do with AI, whether your state bar has said so explicitly or not. For NJ solo and small-firm attorneys operating without a general counsel or ethics compliance team, the Opinion is both a warning and a roadmap.

Read it. Then act on it.