More Than Just Hallucinations: The Unspoken Misconduct Risk of AI Under NJ RPC 8.4

Most conversations I have with New Jersey solo and small-firm attorneys about AI ethics revolve around two familiar fears: confidentiality breaches under RPC 1.6 and inaccurate 'hallucinations' that could lead to sanctions under RPC 3.3. While these are critical concerns, they obscure a more fundamental risk—one that strikes at the core of our professional integrity.

The unexamined rush to adopt AI for efficiency's sake is creating a minefield under NJ RPC 8.4, which governs professional misconduct. Specifically, subsection (c) states it is professional misconduct for a lawyer to "engage in conduct involving dishonesty, fraud, deceit or misrepresentation."

We tend to think of this rule in terms of overt deception. But in the age of generative AI, misrepresentation can be far more subtle. It's not about what the AI says, but what your use of it implies.

The Misrepresentation of Provenance

When you send a client an email, file a pleading, or submit a brief, you are making an implicit professional representation. You are communicating that the work product is the result of your legal training, diligence, and professional judgment. What happens when it isn't?

Consider this common scenario: A client emails with a complex question. To save time, you feed their query into an AI, which generates a comprehensive, well-written response. You give it a quick read-through and send it under your name. There are no factual errors. No confidential data was breached. Have you done anything wrong?

Potentially, yes. You have misrepresented the provenance of the legal analysis. The client believes they are receiving a bespoke answer crafted by their trusted counsel. Instead, they received a machine-generated text with light human oversight. This isn't a matter of competence under RPC 1.1; it's a matter of candor. The efficiency gain comes at the cost of a small, implicit deceit—a papercut that can erode the foundation of the attorney-client relationship and, arguably, cross the line into a misrepresentation under RPC 8.4(c).

When Automation Becomes Deceit

The risk escalates when these AI shortcuts are systemized. Let's look at two other areas ripe for RPC 8.4 violations:

1. Automated Client Intake and 'Evaluations': Many firms are deploying AI chatbots on their websites to engage potential clients. If your bot is programmed to say, "One of our attorneys will review your submission and get back to you," but the follow-up is an AI-generated summary and a generic call-to-action, you are being deceptive. The potential client is led to believe an attorney has personally assessed their matter. This is a clear misrepresentation of your services, moving beyond a simple advertising violation (RPC 7.1) into the territory of deceit.

2. AI-Assisted Discovery: Using AI to draft initial responses to interrogatories or document requests can be a powerful time-saver. However, the supervising attorney's review must be substantive. If a partner performs a cursory review of AI-drafted responses and signs the certification, they are representing to the court and opposing counsel that they have diligently and personally ensured the accuracy and completeness of the responses. If that review was merely a formality, and the attorney was fundamentally relying on the machine's output, it could be argued they have misrepresented the extent of their professional involvement in a crucial stage of litigation.

How NJ Firms Can Practice Principled AI Adoption

Avoiding these ethical traps doesn't mean abandoning AI. It means integrating it with integrity. For solo and small-firm practitioners in New Jersey, here are three practical guardrails:

• Internal Labeling, External Judgment: Implement a firm policy that requires any AI-generated text used internally—from research memos to draft emails—to be clearly labeled. This creates a culture of awareness. For external communications, the standard is higher: the final work product must substantively reflect the attorney's own analysis and judgment. If the AI's contribution was significant, the attorney must essentially "adopt" the work by thinking it through, not just proofreading it.

• Audit Your Automation for Implied Promises: Review every client-facing automated system, from your website chatbot to your email auto-responders. Scrutinize the language. Replace phrases like "our team will review" with more accurate descriptions like "our automated system will process your information to determine if we can help." Transparency builds trust; implied promises that you can't keep destroy it.

• Redefine Your Role as 'Chief Validation Officer': Your primary value is not writing the first draft; it's your professional judgment. Frame your use of AI accordingly. You are not a glorified proofreader for a machine. You are the final human validator, responsible for every word, every argument, and every implication. This mindset shift is the single most important defense against inadvertently violating RPC 8.4.

Ultimately, the ethical integration of AI is less about the technology and more about our professional character. By focusing on honesty and transparently representing the nature of our work, we can harness the power of AI without compromising the duties that define us as New Jersey attorneys.