- law dawg
- 35 Comments
- 2762 Views
Go check out this story at Original Jurisdiction titled, “A Major Law Firm’s ChatGPT Fail.” Where do you fall on the use of AI in the legal profession? We know some of you are using it (shout out to whoever used it to summarize the bench bar meeting yesterday) and we suspect some of you aren’t even sure what it is. For those of you using, what kind of use cases have you discovered? Have you seen it cite “hallucinated” cases? Do you see AI as a practical tool that more people should be using or are you cautious? For those of you not using, why not? Will you ever?
AI is useless, I’ve tried it before. It absolutely makes up cases that don’t exist.
I am the person who posted the AI summary yesterday. You are welcome.
I use AI all the time, every day. It has very obvious limits. You cannot, for example, just dump a bunch of documents and notes into a black box and have it spit out anything that is coherent, much less a final product, at least in litigation. However, AI is extremely useful in smaller, discreet steps during a project. For example, in IRAC, AI is terrible at I and R, but pretty good at A and C (although even then some fine tuning and editing is needed).
I use AI to help develop strategy and tactics. Right now, I feel that I have a bit of an edge compared to most of my colleagues. Soon, however, that will disappear. For now, I am careful about sharing what I have learned because I would like to enjoy this temporary advantage.
One of my fears is that AI will homogenize writing tone, voice and style. I can feel AI having that effect on me. I am trying to keep up my non-legal recreational reading to keep my writing voice authentic.
Also, AI can reveal writing weaknesses. I *think* I’m a good writer, but it’s in a real Uncle Rico kind of way since I’ve been at this for a while. AI has been a good wake up call that I have become complacent and need to tighten things up and be more thoughtful ad deliberate about how I structure sentences, paragraphs and arguments.
Anyway, all that said, 9:21 AM is correct. AI is useless. Don’t try it. You’ll be just fine. 😉
You don’t say. ” I HOPE THIS MOTION FINDS YOU WELL…”
I find this fascinating and want more tips
Reading these comments makes me glad I will be retired in a year or two. Everyone should be required to learn how to Shepardize with books. Harrumph.
When it works it’s amazing. I can feed the best AI models a 100 page pdf and get insightful, detailed answers to specific legal questions (i.e. provide a complete timeline of the relevant events giving rise to this litigation, evaluate the strength of the arguments for and against the relief sought by the parties in this motion and opposition, etc.). The problem is that the answer is only 85-90% accurate. Figuring out what parts are correct and what parts are incorrect requires me to read the entire 100 pages, meaning I didn’t save any actual time at all. Fancy AI legal research tools all have the same problem. If I have to do all the research myself to ensure nothing is missed or wrong, why bother?
Claude is pretty good at identifying actual cases. I’ve actually discovered a few cases here and there that ended up being very helpful, or pointed me in the right direction. However, you have to be careful. Often times AI will cite to real cases, but will completely misconstrue the holding by the court. You have to read through the cases it cites entirely so you don’t end up looking like a fool.
AI is great for preparing outlines. Claude let’s you attach files you wish to center its answers on. Avoid using documents with sensitive information, or switch out names and identifying information. If you’re stuck with how you want to structure a Motion or letter, just throw your materials into Claude and ask it to provide an outline.
Also, if you’re using the free version of any AI, you get what you pay for (nothing). Pony up the $20, you’ll notice an immediate difference.
If you’re a bad prompt writer, you’re gonna be disappointed. Garbage in, garbage out. You should be provide your AI with a full paragraph or two of context before you start jumping down its throat for answers. Don’t just expect it to create magic after you feed it 2 sentences of vague nonsense.
I personally don’t use AI to change my writing style. Feels too sterile, and a person familiar with AI will see right through it.
9:49 here. One more thing: AI is FANTASTIC at answering questions related to NRS, NAC, or court rules. Those sources are much more digestible for the AI in comparison to case law. You can generally trust what the AI is providing you with regards to statutes, regulations, and rules.
The more AI introduced, the less useful anything is. This is true for anything from predictive text to Westlaw.
Hell, the Westlaw searches now are no different from grabbing a case in the dark and hoping its correct, even with Boolean searches. We are on the testing team for Westlaw AI (new changes), and I have to tell you –it is the worst thing that has ever happened to me. Even the flags are wrong. Too many cases are flagged as red or overturned when you look, and see the case is cited favorably. That’s absurd. Westlaw’s wannabe ChatGPT is even worse. You ask it a question, and it gives you wrong answers.
Google is effectively defunct, as it cannot return a basic response to a basic search.
ChatGPT has only ever been useful for finding recipes. Anything deeper is a shitshow.
Even Grammarly has it flaws. First, stop correcting the grammar in a quote, we have to leave it that way. Second, the word “the” does not have to be in front of defined terms. Third, rephrasing something to make it say something completely different is not helpful.
AI is an absolute nightmare. They need to perfect it much more before even thinking of rolling it out.
AI is great. I’ve been out of the country for over a month and my office hasn’t even noticed. SKYNET engage.
NPR’s Morning Edition had an interesting piece yesterday about the use of AI in the legal profession, and the expert said we will get to a point where not using AI will constitute malpractice. Fascinating
Who was the expert? Not using AI for what specifically?
It was a woman who was an attorney who wrote a book about the future of licensing for professionals. Didn’t catch her name. And that was a one sentence statement. I’d also like to know more
The expert turned out to be an AI Bot itself
“AI” seems to be a term that every tech firm is using for every software function in 2025. But I’ll breakdown my interaction with AI.
First, AI is absolutely useless as a replacement for legal research. Generative AI like chatGPT will absolutely make up cases and I have had serious issues even with the internal WestLaw AI. I have asked for a certain holding or rule, and the AI will give me a “summary” that is exactly what I am looking for. BUT, when I look at the text of the case the part it cites is actually the argument the court REJECTS, and the holding is the opposite. Admittedly it finds an okay jumping off point for more research, but the non-AI search function provided the same value without the dangerously incorrect summary.
Second, using AI for litigation functions like summarizations of pleadings and discovery comes with similar drawbacks. I feel like every time I get one of these summaries and double check, I find that the AI misses massively important points or even states the opposite of a proffered argument. Just as bad, for something like a MedChron my staff has to spend hours inputting and optimizing data in order for the AI to provide anything usable. In that time, my staff could have just done the damn MedChron! And like always, I simply do not trust anything produced by AI. I always spend the time and effort to double check every detail, and again, with that time and effort I could have done things the old fashioned way.
Finally, I will say that generative AI is pretty good when it comes to legal writing style. By far the most useful function of generative AI that I have found is to put a document in and ask it to edit and punch it up. I have to be careful because it will sometimes change too much, but often I like many of the stylistic editing changes it proposes.
I am a believer that AI can and will provide tremendous value to the legal field. But I have had a very difficult time capturing that value so far.
FYI: No cost bonds for appeals from District Court to Nevada Supreme Court any longer. The bonds were getting lost in the Clerk’s Office and sitting for years and sometimes over a decade. Therefore NRAP amended so that cost bonds are only required if order by court.
“In a civil case, unless an appellant is exempted by law, the district court may require an appellant to file a bond or provide other security in any form and amount necessary to ensure payment of costs on appeal.”
I just received one last month from a case that was 12 years old. i had completely forgotten who the clients were, we had no contact information anymore. It took a fairly significant effort to find the people and they had completely forgotten the $500 and were almost annoyed at being bothered about it.
AI can be pretty good for drafting contracts, the key is you have to give it specific instructions. If you just ask for sales contract, you will get garbage. But if you ask for a sales contract that contains these 10 things… You will get a good starting point at least.
First of all, it’s two letters. It means artificial intelligence, but ultimately what it is, is it’s about machine learning.
We should stop calling large language models AI
https://newrepublic.com/post/191313/donald-trump-ally-supreme-court-overturn-press-protection
Steve Wynn is a massive piece of shit.
I prefer my government being in direct cahoots with the media to fabricate and/or drown out stories!
Come on. Be serious. Can AI take Judges to strip clubs? Pay for dances? Can AI know which Judge ignores which parts of the law? NO!! AI is useless. Look to the billboards for legal expertise.
Yes, AI or LLM like ChatGPT can hallucinate cases, but is really is pretty good at looking at statutes, analyzing strengths and weaknesses, and grammar suggestions. That said, I think of AI as like an associate or paralegal that can do some research (albeit much faster) and give a draft of what we should be doing. Then the responsibility is on me to check the arguments and make sure it makes sense. It is definitely a useful tool in the right hands.
I use AI to review arguments and writing. I never trust it to get anything right but it can shorten some tasks.
Has anyone tried AI in depos? There is a program that apparently records the depo and provides suggestions for follow-up.
I ran into an ethical dilemma this morning. I heard an advertisement for a lawyer that contained an outright lie. “At the end of a case, most lawyers take more money than you.” So the same Bar that wants me to give them $100 every time I boost a Facebook post either doesn’t know about it or has no problem letting that past. So, reportable, but I believe that regulating marketing is BS anyway, and the Bar shouldn’t be doing it. Even if I didn’t believe that, I generally refuse to help DH persecute fellow attorneys unless a client is being harmed.
So, report, or listen to the still small voice inside that tells me to screw Bar Counsel and the rural prosecutor horse he rode in on?
(1) The Holy Ghost said no such thing.
(2) Provide me evidence that this is an objectively falsifiable statement: “At the end of a case, most lawyers take more money than you.”
(3) Which RPC, specifically, has been violated?
Shouldn’t the burden be on the one making the statement to prove its truthfulness?
If the fee is 40% how can a lawyer ever take more than the client?
Let’s say you’re in a car accident and incur $5k in bills. Your attorney incurred $1,000 in hard costs that you have to reimburse. The insurance settles for $10k.
$10k.
-$4k — 40% to the attorney
-$3k — attorney got your medical liens reduced
-$1k — reimburse your attorney’s costs
That leaves $2k for you (after the atty got $4k).
I think most good attorneys would lower their fee to make sure their clients get more than they do (that’s what I’ve seen before). But it’s not outside the realm of possibility that the attorney makes more than the client depending on the fee breakdown and liens/costs.
1. Maybe you weren’t part of the conversation.
2. Do 50% of lawyers take 50% of their clients’ settlements? Does the speaker have evidence of this claim?
3. Nice try, DH/PP/BH. Read your own NRPC.
(1) The Holy Ghost said no such thing
Maybe he did. Cf., Teancum, Nephi, et al.
/s
I hope this post finds you well …. haha. AI phrases will certainly come into the fore and will be a tell for those looking.
I haven’t read all the posts, but
I have found that AI is good at making a general outline, and organizing my thoughts.
AI is not a magic bullet that will solve all your problems. The way I explain it is to think of AI as a trainee or assistant, someone who starts with no knowledge and requires guidance. I used Westlaws AI-powered search engine and found it to be quite effective. However, at the end of the day, the responsibility still falls on you and your professional competence.
For example, if you are some one who fixates on writing style because it doesn’t match your own, AI may not be the right tool for you. But if you value your time and prefer to focus on substantive work while allowing AI to assist with mundane tasks like document editing or support in research, and if you have the patience and knowledge to train your AI platform, then AI can be a valuable asset.