A Word of Caution About AI “Legal Research”
In recent months, we have noticed a growing trend. Many people write to us explaining that they have researched their legal issue using AI tools such as ChatGPT and believe they have discovered cases that support their position. Unfortunately, in many instances the conclusions generated by AI are simply wrong. We regularly see examples where the cases cited do not exist at all, or where the cases do exist but do not say what the AI claims they say. Sometimes the holdings are completely mischaracterized. Other times the AI invents case names, citations, or legal principles that have no basis in reality. This phenomenon is known as an AI hallucination, and it has already caused serious problems in real court proceedings when lawyers or litigants relied on fabricated case law.
We understand that hiring an attorney can be expensive. Legal representation is not always financially possible for everyone, and that reality forces some individuals to consider filing cases pro se (representing themselves). However, aside from the general bias against pro se litigants, there are many reason why we generally discourage people from proceeding pro se whenever possible. The legal system is complex, and small errors in legal research or procedure can have significant consequences. The rise of AI-generated legal research unfortunately gives us another reason for that caution. If you are using AI tools to help guide your research, please treat the output as a starting point, not a final answer.
We also understand that most of you don’t have access to professional legal research databases such as LexisNexis or Westlaw. But you still have access to a powerful and free resource: Google Scholar. By going to scholar.google.com and selecting the “case law” option, you can quickly verify whether the cases you believe support your position actually exist and what they actually say. You do not need to be well-versed in legal research to perform this basic cross-check. If a case cannot easily be found there, that is often a strong indication that something is wrong with the citation you were given.
AI tools can be helpful when used carefully. But relying on them blindly can lead people down the wrong path. Worse, it can expose you to sanctions (even ProSe litigants can be sanctioned for this) and you risk creating bad precedent which will be used against someone who later brings a case with merit and proper citations.
Discover more from Florida Action Committee (FAC)
Subscribe to get the latest posts sent to your email.
I’ve had a handful of folks already spout AI “research” to me. I wrote a piece about this in my newsletter last year. AI still gets things wrong more often than it gets right. Count me among the AI haters, if you will. It is bad enough having to deal with the folks who think they have innovative legal solutions to our plight without adding those who got ideas from a computer algorithm. If anything, these folks are making my job harder. I get it, we want “easy” and “quick” answers but that line of thinking caused the laws to become entrenched in our society already.
At the least, fact check AI slop and don’t blindly accept results without verifying what is said.
Because this sentence will be largely overlooked by AI haters and people looking to blame AI for their troubles I want to make sure it is seen – “AI tools can be helpful when used carefully”
What that means is results AI kicks out should be read, understood and verified. Does this mean you need a professional to do all of that? No. But if AI kicks out something that fits your belief or narrative, question it. Ask for sources “where did you get that information?” and challenge AI to defend it’s findings. AI slop can compound, but when used efficiently it can provide a great deal of value.
It would be nice to see an article that speaks to the “how” to use AI responsibly instead of just warning of the dangers.
I use AI extensively for a number of things and I always validate my work either myself or through having it reviewed by folks with eagle eyes for the subtitle nuances that can become larger problems. This really is no different than someone misinterpreting the law or a legal case and writing their own paper with improper findings. It’s just more efficient at doing the wrong things as much as the right things.