Does Giving Your Paralegal Generative AI Create a UPL Problem Under NJ RPC 5.5?

You’ve invested in a generative AI platform to make your New Jersey practice more efficient. The next logical step is to give your star paralegal access. They can summarize depositions, organize discovery, and draft initial communications in a fraction of the time it used to take. What could possibly go wrong?

Unfortunately, quite a lot. The moment a non-lawyer staff member uses an AI tool to perform tasks that require independent legal judgment, your firm is wandering into the territory of the unauthorized practice of law (UPL). This is a direct risk under New Jersey’s Rule of Professional Conduct 5.5. More importantly, it’s a risk that lands squarely on your shoulders as the supervising attorney under RPC 5.1.

Where Information Retrieval Ends and Legal Judgment Begins

The core of the UPL rule is about preventing non-lawyers from exercising the professional judgment reserved for those admitted to the bar. With AI, the line is defined by the user’s prompt. The same tool can be used for permissible administrative tasks or for activities that constitute the practice of law. The distinction is critical.

Consider these two prompts given by a paralegal to an AI:

• Permissible Task: “Review the attached 300-page deposition of Jane Doe and create a timeline of every date she mentioned meeting with the plaintiff.” This is a sophisticated data extraction and organization task. It saves hours of manual work but does not require the AI or the paralegal to form a legal opinion.

• UPL Risk Zone: “Analyze this complaint and draft a set of affirmative defenses based on New Jersey case law, prioritizing the strongest arguments.” This prompt asks the AI to weigh legal options, interpret facts through a legal lens, and formulate a legal strategy. When a paralegal directs this work and reviews the output without an attorney’s substantive guidance, they are engaging in an activity that looks a lot like practicing law.

The danger is that modern AI is so capable that it makes the second prompt seem as simple as the first. For a busy paralegal trying to be helpful, asking the AI to 'suggest legal arguments' can feel like a natural extension of research. For a supervising attorney, it's a serious compliance failure.

Your Non-Delegable Duty Under RPC 5.1

It’s tempting to think that a simple verbal instruction to “be careful” or “just use it for research” is enough. It is not. RPC 5.1 requires partners and supervising lawyers to make reasonable efforts to implement measures that give reasonable assurance that the firm’s personnel comply with the Rules of Professional Conduct.

When you introduce a tool that can instantly generate legal analysis, your supervisory obligations increase. You are responsible for the systems your firm uses. If your system allows a non-lawyer to easily generate work product that would be UPL if done manually, you have not met your burden under RPC 5.1. The responsibility to create and enforce these systems is non-delegable.

A Practical Framework to Mitigate UPL Risk

The goal is not to lock non-lawyers out of these powerful efficiency tools. It is to build guardrails that ensure they are used ethically and appropriately. Here are three concrete steps every NJ solo or small firm should take:

1. Develop a “Permissible Use” Policy. Your standard AI policy is not enough. You need a specific addendum for non-lawyer staff that provides clear examples of acceptable and unacceptable tasks and prompts. Prohibit prompts that ask the AI for legal advice, strategy, or outcome predictions. Frame the tool's role as a high-powered assistant for information management, not a junior legal analyst.

2. Institute a “Review and Adopt” Workflow. No substantive AI-generated work product created by a non-lawyer should ever be used, filed, or sent to a client without an attorney’s direct review and formal adoption. The attorney must do more than just glance at it. They must scrutinize the reasoning, verify the sources, and make it their own work. This process should be documented, creating a clear record that a licensed NJ attorney made the final legal judgment.

3. Conduct Specific, Scenario-Based Training. Show, don’t just tell. During training, provide your staff with example prompts. Walk through one that is clearly permissible, one that is clearly a UPL violation, and one that is in a gray area. Discuss why the lines are drawn where they are. This builds your team’s institutional knowledge and demonstrates you are taking your supervisory duties seriously.

Giving your team access to AI is a smart business decision. But that access must be paired with a clear, enforceable framework that respects the ethical lines of our profession. Your AI policy is incomplete until it directly addresses the UPL risk and establishes the specific supervisory checks you will perform. This is an active duty, and getting it right protects your clients, your staff, and your license.