Artificial intelligence, or AI, is an interesting new tool for finding authority on which to base a legal argument.  NLRG subscribes to Lexis's AI service, Protégé, and we have several years of experience with using Protégé on a daily basis.

It is critical to understand at the outset that while AI shows promise as a research tool, it can be quite dangerous to the user if it is not used properly.  At its worst, AI "hallucinates"--- creates and cites false authority which does not actually exist.  Over 1500 United States attorneys have been sanctioned for citing hallucinated authority to a court.  The problem is so widespread that courts are charging attorneys with knowledge that the problem exists, even if the attorney tries to claim ignorance.

How, then, can a user obtain the benefits of AI without running the risk of sanctions?  A New York decision provides the answer:

Use of AI is not the problem per se. The problem arises when attorneys abdicate their responsibility to ensure their factual and legal representations to the Court—even if originally sourced from AI—are accurate. . . . When attorneys fail to check their work—--whether AI-generated or not—--they prejudice their clients and do a disservice to the Court and the profession. In sum, counsel's duty of candor to the Court cannot be delegated to a software program.

Ader v. Ader, 2025 NY Slip Op 51563(U),  2025 N.Y. Misc. LEXIS 7848 at *10-11 (emphasis added).

The answer is therefore that after using AI, the user must check the results manually to make certain that the AI did not hallucinate authority.  The attorney who uses AI without double-checking the results is running a serious risk of embarrassment, sanctions, and even potential disbarment, if the AI product includes hallucinated authority.

From the moment it first adopted AI, NLRG has focused very directly on using AI properly.  Each research inquiry we receive is assigned to a human attorney.  That attorney is response for contacting the client, setting up the case, and doing any work requested.  Our clients are not dealing with a faceless computer; they are dealing with an actual human attorney.

NLRG's policy is that AI is one among many tools which our attorneys use in preparing products.  Our attorneys use AI when it is likely to be helpful; they do not use AI when it will not be helpful.  When our attorneys are AI, we require them to check the results manually.  This practice ensures that we do not cite hallucinated authority to a client.

But the problem extends beyond the outright hallucination of authority.  Even when it is not hallucinating, AI sometimes misreads cases.  It sometimes fails to find on-point authority, especially if the wording of the case differs from the words used in the prompt.  And AI has only a limited ability to helpful analogous authority---authority which supports one or more points essential to an argument, but which stops short of accepting the entire argument.

We have seen cases in which a simple AI search provides useful authority.  Unfortunately, we have also seen cases where an simple AI search went down the wrong path, missed relevant authority, or simply failed to understand the issue---even when no outright hallucination was involved.  And we have seen many cases in which a human attorney, using the reasoning ability of the human brain, found useful authority which AI missed entirely.

So our position is that AI is a tool.  We recognize its strengths, and we use it as one way of finding helpful authority.  But we also recognize its weakness.  We never rely upon it as our only research tool; we have seen too many cases where an AI search missed relevant authority.  We also use word searches, shepardizing, treaties, articles, and other research tools, many of which have been around much longer than AI.  The best results are obtained by using all of these tools together---not by using any one tool in isolation.

While we use AI as a research too, we do not use AI for drafting.  We have tested the drafting ability, not only of Protege, but also of other advanced AI tools.  We have consistently found that AI cannot draft a brief which meets our quality standards---standards which we have following since our founding in 1969.  The fact that AI has value as a research tool does not mean that AI can draft a product all by itself.

Because our research attorneys have been using AI regularly for several years now, we have as much experience using this technology as anyone.  By hiring NLRG, you the benefit of both modern technology and traditional methods, together in a single package aimed at finding the best possible authority within a cost you can afford.