Here are 3 examples (there are more): https://github.com/harveyai/harvey-labs/issues/191
I don't know the other fields well enough, but if all your clients end up needlessly detained or admit to things they have a valid defense for, eventually people will notice. It's not exactly the two major sins that will definitely get you disbarred - stealing money and screwing client while representing them - but it's definitely at least incompetent. They seem to not understand that common law jurisdictions are fundamentally different from civil ones, and also, there's quite a bit of difference between how the system evolved in each country. Or that in China what's printed is basically more like a manifesto, since it's almost entirely run by the bureaucracy and personal relationships, and lawyers that aren't foreign-trained and work for foreign firms aren't considered prestigious since the courts are essentially entirely for show.
They need to get like, one jurisdiction, one area in that jurisdiction, and specifically the kind of sub-sections in that area down pat. But they can't train on real material so they will never be able to do so. Their banner testimonials come from non-lawyers and that says it all.
I genuinely appreciate efforts to make access to justice more equitable & cost-effective. And I don’t actually much care if people are using these tools to draft boilerplate contract provisions, create initial document summaries, and other such mundane tasks.
But damn, uncritical reliance on these tools is setting yourself up for potentially serious downstream pain. I’ve yet to see legal automation (including LLMs) that can one-shot situationally appropriate and fully trustworthy outcomes in the absence of experienced human review.
Law is an imperfect effort to codify ever-evolving norms, and interpretation of the law is often more of an art than a science, requiring intuitive leaps and judgements not necessarily suited to statistical prediction machines. And even where the law appears to be reasonably settled, not every legal situation gets resolved in the form of words that somehow end up in a corpus of training data subsequently used to train a model.
In other words: legal LLMs are still very much caveat emptor. Trust if you must, but please, please, please, carefully verify before making potentially deeply consequential decisions based on the output of these tools.