A Hypothetical NJ Disciplinary Case: The Solo Attorney, the AI Tool, and the Missed Conflict of Interest

We all know the story of the lawyer sanctioned for citing fake cases generated by a public AI. But the more insidious risks of artificial intelligence aren't in the spectacular failures; they're in the silent, background processes that our ethical rules were never designed to anticipate. Let’s walk through a hypothetical—but entirely plausible—scenario that should serve as a wake-up call for every solo and small firm attorney in New Jersey.

The Setup: Two Seemingly Unrelated Matters

Meet “David,” a sharp solo practitioner in Bergen County. He’s tech-savvy and uses a popular, commercially available AI chatbot to accelerate his research and drafting. He believes he’s being careful, using it for ideas and structure but always verifying the output.

Matter A: David represents a software developer in a contentious separation from her business partner. The core of the dispute involves proprietary source code. To prepare his arguments, David feeds confidential details about the software’s architecture and unique algorithms into the AI to help him summarize the technical points and find analogous trade secret cases. The representation concludes successfully.

Matter B (Six Months Later): A promising new client, a local tech startup, retains David. They want to sue a competitor for patent infringement. The competitor’s product, they claim, uses a technical process that is remarkably similar to their patented method. David runs his standard conflicts check. The name of the defendant company doesn’t ring a bell, nor do the names of its principals. The check comes back clean. He takes the case.

Here’s the problem: The lead engineer at the defendant company is the former business partner of his client from Matter A. But since that individual isn't a named party and David's conflict software only scans for party names, the connection is missed. The two matters are, in the language of the Rules of Professional Conduct, “substantially related.”

The Breach: How AI Became an Unwitting Accomplice

David dives into Matter B, once again leveraging his go-to AI assistant to draft discovery requests and outline deposition questions. Unbeknownst to him, the consumer-grade AI he used in Matter A had a default data policy: user inputs are retained to train and improve the model. The confidential technical details from his first client’s case were ingested, anonymized, and integrated into the model’s vast knowledge base.

When David prompts the AI about Matter B, the model—its neural network now subtly shaped by the confidential data from Matter A—produces an uncannily insightful line of questioning. It suggests probing a specific, obscure vulnerability in the defendant’s software architecture. It’s the smoking gun.

David, impressed, incorporates this into his strategy. But when opposing counsel sees the discovery request, their client—the former partner from Matter A—recognizes his own proprietary information being used against him. He immediately knows where it came from. A motion to disqualify David is filed, and the disciplinary committee is notified.

The Ethical Violations: A Cascade of Failures

David’s single failure—using a non-secure AI tool—triggered a cascade of ethical violations under the New Jersey Rules of Professional Conduct:

• NJ RPC 1.6 (Confidentiality of Information): The first domino to fall. By inputting confidential client information into a tool that uses inputs for training, he failed to protect information relating to the representation of a client.
• NJ RPC 1.9 (Duties to Former Clients): This is the core of the disqualification. David represented a new client in a substantially related matter whose interests were materially adverse to his former client. Worse, he used information relating to the prior representation to the disadvantage of the former client. The fact that an AI acted as the intermediary is no defense.
• NJ RPC 1.7 (Conflict of Interest: Current Clients): While the primary issue is with a former client, the AI’s data retention created a latent, systemic conflict risk for his entire practice, jeopardizing duties of loyalty.
• NJ RPC 1.1 (Competence): The duty of competence includes technological competence. David failed to understand the terms of service, data privacy policy, and fundamental architecture of a tool he was using for sensitive legal work.

Four Steps to Avoid This AI-Driven Conflict Catastrophe

This scenario is avoidable. Protecting your practice requires moving beyond just checking an AI’s output for hallucinations and scrutinizing the tool’s underlying infrastructure.

1. Mandate Zero-Retention Policies. For any client-related work, use only legal-specific AI platforms that contractually guarantee a “zero-retention” or “zero-inference” data policy. This ensures your prompts and client data are never stored or used to train the model.
2. Scrutinize Vendor Contracts. Your vendor agreement must explicitly state that your firm retains ownership of all data, that your inputs will be kept confidential, and that they will not be used for any purpose other than generating your output. If the vendor won’t agree to this in writing, walk away.
3. Expand Your Conflicts Check. Your intake process should now include questions about the subject matter and key non-party individuals involved. A simple party name search is no longer sufficient to uncover these kinds of “hidden” relationships.
4. Establish a Firm-Wide “Approved Tools” Policy. Even for a solo, you must formally decide which tools are safe for which tasks. Create a simple policy outlining approved AI vendors and expressly forbidding the use of consumer-grade tools (like the free versions of ChatGPT, Gemini, etc.) for any work touching on client matters.

AI doesn't change our core ethical duties, but it creates novel pathways to violate them. The duty of loyalty is absolute, and a failure to understand the technology you use is no longer a viable excuse.