AI and the Law – Friend or Foe?

More and more I have clients send me draft documents, such as legal opinions and letters to counsel, which they have drafted using AI (ChatGBT and the like). For many, I assume this is to help build their case, share their view with me in an efficient manner, and of course, to reduce time and cost. However, in many cases, I find the use of AI counter productive, and below are some of the reasons why.

Before criticizing what many may deem the best thing since sliced bread, I will admit that I do appreciate the advancement of technology and regularly use AI in my legal practice – mostly to speed up my research process. It can revolutionize the law. However, AI and the law have yet to become best friends, at least in my view. A very discerning eye needs to review AI’s work and those who simply accept AI’s legal work product or answer without question may make grave errors based on misinformation, ultimately impacting their life in some way.  

As many of you may have already heard, AI has been known to “hallucinate” and fabricate caselaw supporting the drafter’s position. While I first assumed this would be an uncommon occurrence, I have since had several clients in the last year send me caselaw which, according to all legal databases and public records I could search, did not exist.

Further, and more frequently, clients send me AI driven legal opinions or submissions with caselaw which misrepresents the referenced court decision or status of the law. This has taken many forms, from citing a case that supports a principle which it plainly does not, to misconstruing a legal doctrine, or over generalizing. However, a lot of law is highly fact-specific and decisions can be often easily distinguished from one another, even if they discuss the same topics. It seems, AI is not suitable for interpreting caselaw with a very high degree of accuracy (at least for now).

The result? I end up having to spend billable time checking the caselaw sent by clients which I am not familiar with, very often to find that it does not exist or, does not contain the principle it is relying on or concluding that AI’s draft is a gross misrepresentation of the position it intends to support. Additionally, a client may then come in with greater expectations of their case than those they would have had if they had spoken to me prior to reviewing the AI opinion. I then have to communicate the bad news that their case is not as strong or as clear as they think.

That being said, I use a legal database which has an AI research function that only searches within its own database of caselaw, encyclopedia, memorandum, court pleadings, and the like. It links all of its sources underneath every statement made. Thus, it removes the concern over hallucination. While its summarizing function still produces some misleading results, it is very easy to verify the information and is often a wonderful starting point with locating a needed line of caselaw or that “smoking gun” we lawyers all dream of. As a lawyer, AI is my friend, but it’s not the type of friend I would trust with my life.  

Often, I find clients just want to hear what they want, and AI can be good at that. However, it does not serve an employee to go into a negotiation or costly litigation with their employer without a fully informed decision or the right expectations. Likewise, it does not serve an employer to draft policies or contracts that are not specific to their needs or compliant with their specific jurisdictions and industries. It certainly does not serve them to believe they have the right to act in a certain way in an employment relationship without the often-needed deep analysis of the facts and evidence available. Could these things be resolved by inputting more detailed information and posing the right question to AI? Yes. Will the drafter know what information is relevant and what the perfect question to ask is? Maybe not.

At risk of this coming off as a plug to keep me in business, what I really intend on sharing is my professional experience with AI and the law, and how clients can be mindful of it’s (current) limitations with respect to law. I have no doubt that things will be improved in future and can only hope to be so experienced, invaluable or retired at that point that, I too, am not replaced by AI.