For what reasons would lawyers hesitate to use a legal AI product?

But beyond making things up, legal AI also still struggles with basic research tasks, specifically treating specific rules as universal or exceptions as the rule itself. In the last six months or so (this is now May 2025), I’ve had two clients - both worshippers at the church of AI - freak out that I had allegedly gotten something “very wrong.” Why? They’d asked questions of AI and the answers they’d gotten were 180 degrees different from mine. In one case (in Texas), AI had told them that California law effectively applied nationwide, and in the other (federal), AI had told them about a limited state court exception (and not even the state that we were in!) being the general rule. This was what I ultimately determined after investigating where AI had gone wrong, but the AI answers themselves didn't contain any nuance: in the first case, for example, it didn't say “here is the rule in California,” something that might have clued a reader to the fact that the answer did not apply to Texas. Rather, it said “here is the rule, period.” The AI answer was worded in a way that was so definitive and seemingly authoritative that it would have been easy to assume it was correct.

I think that the progress made by AI - even just in the last year or two - in the legal field is significant, and will no doubt improve over time. But right now, AI is kind of like a really hard-working secretary that has been around the law a lot, but still didn’t go to law school and therefore makes mistakes that a lawyer would not. There’s nothing wrong with using legal AI as a tool, but just as I don’t rely on my assistant to complete complex tasks and double-check her work, nothing AI renders at this point should be regarded as definitive.

Because they’re often extraordinarily wrong. Everyone’s read about the situations of lawyers filing AI-drafted motions that cite to cases that don’t exist. That’s colossally embarrassing, but even before AI, it was always a necessity for lawyers to cite check others’ work to make sure that the cases in their briefs are accurate and still good law. In other words, this is lawyer laziness in the same way that it would be to (a) allow a junior associate you’ve never worked with to draft major motions and not check their work or (b) pull case citations from a 20-year-old treatise without checking to see if the law has changed since publication.

What is the difference in effectiveness between an infrared sauna blanket and a regular one?