As someone with no expertise in legal, the first thought that comes to mind is that law changes.
For a field that is (appearingly) so complex and multi-faceted, surely agentic AI is bound to hallucinate, particularly when components of the law change?
This is really an open question. If anyone with more experience has thoughts on this I'd be very interested.
I don't think it can expert in all jurisdictions and all area of law.
There are lots County or state specific law.
Some case law in different state can be referenced for some legal principle, even if the same law does not exist your etc.
law moves much slower and is much less complex than code. it isn't like computer code though. the interpretations and rulings are mostly fugazi to justify power relations and corruption, so I doubt that an LLM can grok law the way it can computer code, because if you think that logic and precedent actually matter, you are missing the point.
most law firms are basically assembly line processes that have one fixed specialty, with a small set of claims that they will file, and typically targeted at insurance companies, so there is very little real practice of law. you just set up your assembly line to scam insurance in your particular niche and then search for clients that fit your cut out. LLMs can definitely disrupt this model in interesting ways.
For a field that is (appearingly) so complex and multi-faceted, surely agentic AI is bound to hallucinate, particularly when components of the law change?
This is really an open question. If anyone with more experience has thoughts on this I'd be very interested.