Legal AI Ethics Crisis: Lawyer's Use of ChatGPT Leads to Courtroom Ban

0:00Recently, a lawyer used ChatGPT to help write a document and he got in big trouble because ChatGPT made up a bunch of cases that it cited. So this has, I feel like, left the overall law industry kind of reeling and re-evaluating their usage of ChatGPT and AI in general. And there's been some new laws that have recently been enacted as of just a couple days ago because of this. So today on the podcast, we are going to dive into what is happening in the law field with AI. And a lot of this actually stems back to OpenAI itself and also Sequoia Capital, one of the biggest venture capital firms in the United States.

0:40back in a company called Harvey AI. Now, Harvey AI kind of dubbed itself as, you know, the, you know, paving the way for how legal AI works. And essentially, they're saying that they're, you know, training AI models for lawyers, but also they're saying that they're built on GPT-4. So I'm not 100 % sure if you go to their website, all you're going to see is join the waitlist and or, you know, contact sales. If you go to their blog, there's just a couple announcements about them getting money. And if you go to their careers page, it looks like they're opening, they are looking for senior researchers, some backend people.

1:19Honestly, it looks like they had an idea and then got a bunch of money and now they're gonna hire people. So in my opinion, that's where they're going at it. With Harvey AI, they might be further ahead than I believe. But that being said, that doesn't mean that that hasn't stopped over 15 ,000 lawyers from joining that wait list. So a lot of different law firms are really anxious to get their hands on AI. And it makes sense. I can really see the use cases for generative AI. And especially when you have someone like OpenAI themselves backing this company and helping fund the company, you can see why lawyers would be trusting and believing that this is a good way to go.

1:57Now, that being said, I think there's probably a lot of misinformation or people are not understanding correctly the limitations of ChatGPT and OpenAI in general. And that is exactly what happened just recently to a lawyer in particular by the name of Roberto. Oh, sorry. His client was Roberto Malta, who was suing a Colombian airline last year. Essentially, the dude said that a metal food cart and beverage cart injured his knee on his flight to Kennedy International Airport. And so he wanted to sue Avianca Airlines. Airlines a Manhattan judge was asked by the airline to dismiss the lawsuit based on the statute of limitations blah blah blah his lawyer Stephen A.

2:43Schwartz this is the guy in question he submitted a brief based on research done by ChatGPT and he sent it over to the airline's law firm now the airline's law firm was really confused because it pulled up the brief and they couldn't sort or they couldn't cite any of the any of the law or any any of the cases in there. So pretty much there was one called Martinez versus Delta Airlines, Zickerman versus Korea Airlines, Vargas versus South China Southern Airlines. So and not only was he citing sources, but he was citing made up decisions by those sources. So you can you can obviously just assume what prompt he said he said like hey i need to come up with a brief um based off of my client who got injured by a food cart on you know x y and z airline please come up with like some examples of uh other cases that have favorably settled in the case of the client whatever right and then it just made up three cases um and i've had it do similar stuff honestly more than so in the past with gpt 3.5 so i wonder if this guy even knew he needed to pay 20 bucks to upgrade that'd be pretty sad if this wasn't even GPT-4.

3:59But in any case, it made up some cases and this guy did not fact check them and he submitted them. Now, he obviously is coming under a lot of fire. This made like national news and everyone was covering it because it's pretty outrageous that, you know, a lawyer is using chat GPT to write their briefs and not fact checking it and getting caught in the process and so uh what seems interesting uh avianca the the airline their law firm said it seemed clear when we didn't recognize any of the cases in their opposition brief that something was amiss we figured it was some sort of chatbot of some sort anyway so they go talk to the judge judge gets mad at him and schwartz then has to respond in an affidavit because now he's kind of busted and he says that he quote-unquote consulted chat gpt to quote-unquote supplement his legal research yeah right he literally just got it too this is this is not uh you know this is for educational purposes only my opinion on this is that he did not do any research and he uh just made a prompt and copy and pasted what chad gpt said but don't sue me schwartz in any case this is not a what libel you know just my opinion any case um he said that as so pretty much after it gave him the, after quoted those like fake studies, he asked it if they were real and he asked it for a source.

5:25And Chachi Bouttee apparently said that it confirmed that it was a legitimate thing. And it said that the, that he could go find the case cited in the Westlaw and Lexis Nexus databases. So, you know, it's, it's cited real databases. Now what he should have done is went to the databases and looked it up. I don't know who in the world would just trust, you know, something and use it legally and not like just do the research of looking it up. I mean, it literally cited where it said it was getting it. One of them, not all three. So even in this case that he's like, you know, trying to defend himself by saying it cited one, he didn't say that it cited the other two.

6:06So he still submitted to that. Anyways, it's just ridiculous. So pretty much he got busted. He's in big trouble. It'll be interesting to see what happens there. The reason I bring it up is because I think this is going to have a bit of a chilling effect on the law industry in general, right? We have Harvey AI that just recently, apparently it says they just raised$21 million in fresh investor cash. Sequoia Capital was leading the Series A round. And they said that's because more than not just 15 ,000 lawyers, but actually 15 ,000 law firms were on the waiting list to use Harvey. And so OpenAI's startup fund, Silicon Valley Angel, Elad Gill also participated in the funding round.

6:51And so yeah, they're raising like a lot of money. I don't know if they have their product launched yet, or if it's just that. But I think a lot of the hype came when GPT-4 was released, and researchers used GPT-4 to pass the bar exam. So obviously, it has a lot of capabilities. I'm not sure if this lawyer that submitted his last document was using GPT 3.5 or four. I'm concerned about that, definitely. But a global law firm, Allen and Overy, said in February that 35 ,000 lawyers and staff were going to be using Harvey to automate some of their document drafting and research. And in March, the accounting firm PricewaterhouseCooper said that it was going to give 4 ,000 legal professionals access to the Harvey platform as well.

7:36So like, you know, you can see 4 ,000 lawyers from this company, 3 ,500 lawyers from another company, these lawyers do not want to get left behind. And I will say this is kind of funny, because I read a report on routers about this. And they said, I'm trying to remember exactly what they said, they said something like, it's a remarkable pace for a profession that was slow to abandon the fax machine. So, you know, obviously taking dig at lawyers being kind of old school. But I think at this point, it's it is definitely an arms race. If you don't want to get left behind as a lawyer, these things, this, you know, ChadGBT, when done right, really can do a lot of research and is a really powerful tool for lawyers.

8:16But that being said, will this replace lawyers? No, it's just supplemental and lawyers need to fact check everything that's coming out of it. And I think what would be even a more powerful tool rather than just asking it to you know come up with cases that are favorable to whatever it's uh looking for i think it'd be better if the lawyer comes up with a case copies the case pastes it in and say like you know write a summary of why uh if in the style of a brief that i need for this type of document explaining why this outcome was like this but they probably should find it themselves i don't know if it's very good at research right now.

8:52Google Bard might be better at research. But in any case, definitely everything needs to be fact checked. So with all of this going on, a new law was just passed by a federal judge in Texas. And he said that all AI generated content needs to be checked, needs to first off be, needs to be noted. So the judge, the lawyers actually have to say if they're using it, and they have to make sure that it was checked by an actual human being. So I think it's going to be pretty interesting to see if that is a rule that is applied to more courtrooms. That's definitely the one that he is, that Judge, I believe, Star's office has said that they are going to be putting forward for anyone that brings any cases to court in their courtroom.

9:41And this is what he specifically said about this. And I think this is interesting because he did give a shout out to some of the great use cases for AI. So it's not like he's just some anti-AI judge. Obviously, he's looking at some of the problems that are happening in the national news right now and does not want that to be repeated for some obvious reasons. But these are the reasons he gave. And I think that they're actually fairly compelling. He says, these platforms are incredibly power and powerful and have many uses in the law, form divorces, discovery requests, suggested errors and documents, anticipated questions at oral arguments.

10:14But legal briefings is not one of them. Here's why. So legal briefings, obviously, that was the problem that got that original lawyer in trouble in the airline lawsuit. So anyways, the judge goes on to say these platforms in their current state are prone to hallucinations and bias on hallucinations. they make up stuff even quotes and citations another issue is reliability or bias while attorneys swear an oath to set aside their personal prejudices biases and beliefs to faithfully uphold the law and represent their clients generative artificial intelligence is a product of programming devised by humans who did not have to swear such an oath so i think this is actually a pretty compelling um case to make right obviously um it's not just like chat gpt didn't have to swear an oath.

10:56But it's like, also, there's humans that are developing these tools and that are, you know, making it say certain things or have certain ideologies that they believe in. And they did not have to swear an oath to uphold the law or anything else. And, you know, chat GPT might be the big one right now that everyone can pick out. But there's going to be a lot of these different ones. And it's without good transparency, it's really hard to know what the biases have been added to those platforms and if those are detrimental, right? Like, I mean, let me just throw like a crazy conspiracy theory or like worst case scenario that a bad actor could use.

11:31So let's say I'm China and I would like to have a negative impact on the law field in America. So I make a really powerful, like I go buy Harvey, for example, you know, all the law firms are using it. It's the number one. I get like some shadow investors. We go buy up Harvey and then we go and say, you know, for x y and z reasons to be more safe and effective we need um to put these certain ideologies into the into um the into the biases of the ai model and push certain narratives and so all of a sudden every lawyer that's using this is now getting like some sort of you know slant that perhaps is more positive to the chinese communist party or something i know that sounds crazy but like, come on, let's think about it.

12:15Like these algorithms and these AIs, all of this is possible. So I think that we do just want to protect against it and to make sure that, yeah, that still may happen, to be honest. That would be interesting. It's not like it hasn't happened in other areas in the world and in capitalism and corporations and whatnot. So in any case, I think this is one issue that we do need to be aware of as far as the biases of these. Anyways, the one judge, though, he did have one other thing to say. He said, as such, these systems hold no allegiance to any client, the rule of law or the laws and constitution of the United States, unbound by any sense of duty, honor or justice.

12:52Such programs act according to the computer code rather than convictions, yada, yada. So you get the picture. I think this is gonna be really interesting to see what happens in law. If this entire case and this law that, you know, kind of blew up in this one lawyer's face are going to have a chilling effect on law, probably a good thing that harvey already raised their money because i don't know if they'd be able to get more um at the at the moment but perhaps they will and this is definitely an area we'll continue to watch