A Commentary on “Artificial Intelligence” in the Practice of Law
The emergence and rapid societal adoption of Artificial Intelligence has been fascinating to watch over the past few years. As an attorney my thought process has generally been – bummer for them – whenever I read about corporate staff expecting their cheaper more efficient digital counterparts to replace them. Classic lawyer hubris I suppose.
Over the past year or so, the digital tendrils of artificial intelligence seem to be wriggling their way into the legal community. Every few weeks or so, there seems to be another attorney that gets themselves into hot water for submitting written arguments with so-called “hallucinated cases.”
Hallucinated case, for those unfamiliar, are cases that seem to be made up, or “hallucinated” by AI. The AI will produce a proper citation, page numbers, year adjudicated, reporter, and opinion out of thin air. Given the trust the legal community generally places on its procedure alone, its not entirely unreasonable to think – there’s no way information presented in this format with this depth is fake. Again, classic hubris.
At six years in, I’m considered a comparatively new attorney, so it might merely be my relative proximity to the cautionary tales of law school and clerkships that causes me to chuckle at more seasoned attorneys that may have become more relaxed in their practice – justifiably so – to submit a case that someone (or rather something), convincingly told them supports their argument.
What is interesting is the light that these incidents are beginning to shine into the modern day practice of law. In the first year or two out of school and into the meatgrinder of practice, it is beaten into you over and over again what’s acceptable when looking into cases. Check the case, check the connecting cases, check the connecting cases to the connecting cases, and then hope to your god(s) you didn’t gloss over a footnote. It appears, at least in part, that this practice is not as universal as us young bucks were led to believe. That’s not to say I haven’t had my fill of being chewed out for the lack luster product of my legal research, but I digress.
In 2025 alone, there have been a number of attorney’s standing before the Courts of Maryland, or various disciplinary boards, who bend over backwards to explain why the buck shouldn’t necessarily stop with them. The last three I have personally watched, have attorney’s tossing their law clerks (generally unlicensed law students) or paralegals (generally unlicensed employees) under the bus for hallucinated cases that have found their way into motions or arguments. At least one attorney did all but admit he didn’t actually read the motion that was submitted under his signature.
Maybe its my own nightmares from law school and my first job as an attorney, that has made me obsessed with FRCP Rule 11;[1] but the fact that this phenomenon is occurring at all simply boggles my mind. It certainly seems to take the industry as a whole one more step away from the dazzling perfection of ‘suits’ that so many clients want to compare us to.
What’s even more surprising than one or two careless attorneys; is the fact that it keeps happening. The first guy getting yelled at for it – bad luck; you were caught. But this is happening on an almost regular basis. With the pack mentality of the legal community – you’d think that watching one member eat the toxic fruit – the rest of us would learn quickly not to do the same. This is without even getting in to the potential privilege implications and risks of using AI in our client’s cases – but I imagine we’ll start seeing sanctions hearings regarding that in the coming years.
This is of course, not to say that AI is a cancer on the law. Rather, it’s a “user assumes the risk” situation. I know attorneys who have used AI to help better organize documents, or pull specific pieces of information from massive collections of documents. The benefit is obvious when used cautiously, there is huge potential to save clients thousands of dollars through the lifetime of a case. I myself have used AI to help me when I seem to hit a roadblock in research that I can’t get past, no matter how many synonyms I try to trick the search engine with. AI assisted research can be incredibly effective, and terrifyingly convincing. It’s the convincing aspect where, I imagine, practitioners and their staff run into problems. These cases that don’t exist in the Court Reporters are hallucinated in such detail and depth, that it seems silly not to buy into it. After all, who am I to question the late Justice Ginsberg’s opinion as she discusses the injustices of racial profiling – aside from the fact the case never happened. The problem of course is, as it always has been – you should never sign something you haven’t read. We say it to our clients on an almost daily basis – it’s ironic we’re struggling with it ourselves. I’ve only ever utilized the AI run by WestLaw in my practice, which is advertised as a closed circuit. So in theory, if the the case doesn’t exist WestLaw’s database, the AI can’t use it. In theory. I haven’t personally found WestLaw’s AI to hallucinate cases – yet – but given the court’s waning patience and heightened sensitivity to hallucinated cases; I can’t imagine ever wanting to hang my hat on the words of C-3P0’s Great Grandfather. The chances of being sanctioned for filing a motion untimely because you are stuck researching your argument are generally better than being caught in a misrepresentation to the court.
All of this is to say the inner workings of a lawyers personal practice are rarely thrown into the public light. With any luck, all this recent focus into the black box of the practice of law will force any inadvertent carelessness to right itself – but how many attorneys need to get smacked around before it does, is anyone’s guess.
[1] For those that don’t know; or don’t remember, FRCP Rule 11, is the ‘big one.’ To give a barebones explanation – if an attorney signs something and submits it to the court– you are wholly responsible for everything in it – and the penalties can be severe for fraud or misrepresentation.
