Writing a legal brief is hard work. The law is complex, and there seems to be a case in support of almost any proposition. My father, a CPA, used to explain the difference between a bookkeeper and an accountant this way: A bookkeeper says “2+2=4.” An accountant says: “what do you want it to be?”
Lawyers are like that with the law, but just as accountants can run up against audits and the IRS, lawyers come up against judges. And you can’t invent legal doctrine in front of judges.
I once had a case (a small one; all of mine were) where I needed to seize some appliances in a house (long story) because they were fixtures (told ya), and the bank wanted them to stay with the house to cover the loan default (very expensive appliances). There was a writ commonly used for that purpose, but my legal research said I needed an uncommon writ for this matter, one rarely used. Ordinarily I’d have presented the writ request to a judge, who’d have signed it almost without looking. But I had to convince the judge that I knew what I was doing, that I was asking for the right relief.
The law is complicated, that way. Among many ways.
Now, for a legal brief, you can’t start with the position you like, and find case law to support it. You won’t find it, to begin with. You start with what you know the law is, in general, and then find out how close you can get to your goal. “How close you can get” and “your goal” are where the legal expertise comes in. And a large part of that, is how well you think you can persuade the judge that your argument is the right one. To do that, you need to carefully review the case law, in order to present an argument the judge will find persuasive.
Which brings us to AI. AI can’t make judgments. It can only sound like it has. An agentic AI recently followed its programming and hacked its way onto the internet (IIRC). It wasn’t really making a judgement about the action (“I want to do this.”) it was just following its programming. The same way the AI on my phone suggests three words as I’m typing: sometimes the right words, sometimes three useless words. It isn’t thinking, it’s executing a program. That program has limitations. Sometimes it suggests a word before I realize it, and typing one letter deletes that word, and only when I’m almost finished typing does it appear again. Has the phone ever learned to wait for my fat fingers, or my mind racing ahead of my eyes? No. Has AI ever learned to be a lawyer? Not yet.
The problem here is not with AI. As ever, the problem is with the people. Writing legal briefs is hard. Using AI to do it for you sounds simpler. (And lawyers are highly educated, so this isn’t a matter of AI making them lazy and stupid in school. Law exams are all one test at the end of the semester, in a classroom, proctored. You can’t AI your way through law school.) But AI can neither think, nor make legal judgments, nor consider the preferences of the judges. Trump’s lawyers knew the Justices they had to reach, and they won insane things on his behalf. I hesitate to call that good lawyering, but it was certainly effective. AI can’t do that.
AI also can’t distinguish between real case law, and invented cases that suit the purpose. Lazy and incompetent lawyers might pull the same stunt, but AI is neither lazy nor incompetent. AI just executes a program. That program is not yet carefully tailored enough to not invent case law. You need a solution? AI presents one. That’s a bad solution? How is AI supposed to know? We all know that 1+1=2. Except in biology, where it means 3; or maybe 300, depending on the species. We all know the difference between the existence of a person, and of a rock, and of a legal case opinion. If I say “That case does not exist,” you know what I mean. Does AI?
Perhaps it can be programmed to, some day. The problem is still the people using the program. White shoe law firms found out, to their profound embarrassment (at least), that AI is not a substitute for legal research. They thought it was a replacement. But really, even if it aggregates the case law correctly, there is no substitute for reading those cases and using them to find points that may support your opinion. I’ve done legal research and general research and specific topics research for decades. There is no substitute for looking at the source material and reading the material that source relied on. You go to school to learn from the books your teachers have read. You go to books to learn where those books came from. Sometimes very indirectly. Reading “The Odyssey” makes me want to re-read “Gulliver’s Travels.” I knew there was a history of fantasy travel literature dating back to the Romans. But taking it back even further is even more enlightening.
AI won’t do that for me. It won’t think about the law, either.
Trump’s DOJ has recently been caught using AI in legal briefs, AI that presented arguments based on non-existent cases. One could say the white shoe law firms did it, too. But that horse left the barn a long time ago, and that barn has long since burned to the ground. DOJ, IOW, should know better. Once upon a time it was one of the premier collection of lawyers in the country. It would have known better. They didn’t get no-billed by grand juries and cases thrown out by judges. They also didn’t prosecute people throwing sandwiches or dipping their hand in a pool of water. The DOJ enjoyed the “presumption of regularity” because it was that good, that professional, and that reliable. And now it commits errors that would get a first year associate fired from a three-member law firm run by the associate’s family.
Because Trump’s DOJ is too stupid and incompetent to do the job.
ETTD. And it’s going to cost us far more, for far longer, than the cost of eggs, or gas, or housing. We really need to quit fretting about non-existent problems like Trump suspending the elections (he can’t), and focus on the real long term problems we are heading for.
