Transcription
Mic in the way. All right, welcome everybody to Legal Tech Week for the Friday the 13th edition. Uh, it's Friday, February 13, 2026, and we've got, I think we got a bunch of stories in which AI has, uh, has not been lucky for, uh, for for some, uh, lawyers and maybe even judges, uh, and others. So, we've got a lot to talk about today. Um, this is the show, if you haven't been here before, where we talk about the top stories of the weekend, legal, tech, and innovation. I'm Bob Ambrosie. I have a blog called Law Sites and a podcast called Law Next. And our panelist today, uh, I'm going to do something unusual for me, and you don't even know where I start, but I'm going to start at the bottom of my screen, which for some reason, that's me, Victor. Uh, you're you're you're the low man on my screen totem poll today, so we'll start with you.
>> Yeah, that's that's about right. Um, my name is Victor Lee. I am assistant managing editor for the ABA Journal. Um, uh, yeah, I guess I'm, you know, I'm the I'm the bottom of the totem pole. So, you know, that's or I guess what the base of the pyramid or or whatever whatever it is. So, there you go.
>> Yeah, the basic the core the foundation of the Brady Bunch family. Uh, Stephanie, how about you?
>> Um, hi everyone. I'm Stephanie Wilkins. I am the director of content at Legal Tech Hub. And I feel like I should have done my So last week I did the whole um Final Girl horror movie presentation. I should have done that for Friday the 13th.
>> Yeah. All right. Uh Ree.
>> Sure. Um hey guys, I'm Ree Dipshan. I'm the editor and chief of Legal Tech News. And sorry for like the um the sunshine behind me. It's just a halo apparently. Um I'm just on my couch uh this Friday the 13th, but it's awesome to be be here with everybody.
>> Good to have you. And Caroline,
>> yeah, Reese is just showing off the sunshine because it's 8:00 p.m. It's pitch black here like Caroline Hill, editor of Legal Lighty Insider. It's great to be here. Thanks for having me.
>> Great to have you. And Nikki.
>> My name is Nikki Black. I'm the principal legal insight strategist at AM the team behind my case law case and docket wise. I write legal tech columns for ABA Journal, Above the Law, and the Daily Record. And uh um I don't know. I don't have anything else to say.
>> That was enough. That's good. And last but not least, Joe.
>> All right. Before I say my thing, I want to say we need to get somehow whatever it is the rights to that Owl City song Fireflies and just play that as the show begins as everything in the chat is just like Fireflies joins. Fireflies joined. Fireflies joined. Um, all right. Maybe nobody remembers that song. I'm Joe Patrice from Bubble Law. I am, uh, also the the host of Thinking Like a Lawyer. I am, you know, I have bad news for everybody this week. Uh, Microsoft's uh AI chief says that all white collar professional work will be totally automated 100% within 12 to 18 months. So you know if the people behind teams think that then
>> Is that right? Is that a thing? I didn't even say that.
>> Yeah, he said it but hilariously also had said like two years ago that we will have hallucinations completely gone by the end of the six months and stuff like they they're just all completely lying but whatever.
>> Maybe they can just make teams better. There we go. How about that?
>> Make it barely functional. No.
>> Don't you like dread it when you get an invitation to a Teams meeting? It's just like the worst thing.
>> And it's always like it's only like big law firm people who do that.
>> I like I like
>> You like go on
>> use teams. I'm going to just come out and say it. But I like
>> Oh god. So Teams
>> we use Teams too, but we don't have a choice because we we didn't want to pay for the um we didn't want to pay for the Zoom license anymore. So you know Teams is Teams is free with your Microsoft uh license.
>> Yeah.
>> You know what being invited to
>> You know what being invited to teams and to some extent Google Meet is in the it is the teleconferencing equivalent of saying making your order and them say we have Pepsi. Is that okay? Uh, that is what it is. You see it, you're like, no.
>> Royal Crown Cola.
>> Yeah, that's Google Meet.
>> You don't like Google Meet?
>> I like Google Meet.
>> Well, I also like RC, so I thought
>> Yeah. Yeah. There you go.
>> All right. All right. Um, yeah. Well, uh, so it's sunny where Reese is, dark where Caroline is. It's I, for me, I'm on the West Coast this week, and it's it's still feels like morning, so it's kind of weird doing this show when you feel like it's usually like the end of the week when we do the show, and I still have half a day to go.
>> Yeah.
>> Uh, so we'll see how that goes. Um, all right. Well, we do have a bunch of interesting stories this week, but I think uh I think uh the one, Stephanie, that you picked is probably the good one to start with because it's kind of a case of first impression, as lawyers like to say. It's something that really hasn't been talked about before.
>> Yeah. Um, I the link I'm going to share is to it was actually a law firm blog, but, uh, yeah, it's talking about whether or not AI generated materials can be privileged. Um, this is an issue that I've gotten on stage and talked about multiple times. It's I've been sort of a beating this drum at Relativity Fest for a couple years. So I was glad to see that Judge Raikoff kind of agrees with me here. But, um, yeah, basically came down and ruled that, um, it unfortunately there's no written ruling. It's just an oral ruling. But, um, yeah, so I guess a client had used AI to create something that he then passed on to his lawyer and assumed that would be, um, attorney client privilege and surprise, it's not. You can't, I mean, that's not a new doctrine. Like you can't cloak something in privilege just by sending it to your lawyer. We all know that. But he extended that Judge Raikoff extended it in this case to that even though I guess you could argue that the lawyer had generated it specifically for that purpose but denied that and also went further, um, and suggested that, you know, as we've been saying from the beginning, be aware of your confidentiality and your privilege. If you're putting things in third-party tools, you could be breaking that. So that's also a a privilege violation there. And he kind of just threw out all of the arguments a lot of people have been saying, whether it's work product or whether it's attorney claim privilege, which is the boat I've been on for a while, and I don't know if everyone agreed with me, but this is the first ruling of its kind that we've seen anything really definitive in this area and I'm really curious to see what comes next.
>> Yeah, I I thought, uh, I thought Rakeoff was probably right based on where things are, but probably wrong based on where things need to be. Uh, you like like when you when you look at the, uh, briefing, one of the things that they talk about is like, you know, your notes that you create for yourself to have a conversation with your lawyer and like they don't identify what these documents are, but you know, having if you if you're a former litigator, you you know, like it was like creating timelines and stuff like that to he'd already spoken with his lawyers. He then created these documents and then gave them to his lawyer. Uh, so it's like what the timeline was and stuff like that. Those are very much the sorts of documents that All right, I've got screaming in in the background here, but those are very much the kinds of documents that are normally protected as the confidential notes that you take and then hand on to your lawyer. Uh, and the argument was that that falls apart because he pinged, uh, because he pinged the AI. Uh, which would not have fallen apart if he'd just like done a Google search theoretically, but historically, but he did because it was an AI and that strikes me as a little problematic, uh, given the way people search now. And I also thought that the, uh, the idea that, oh, well, it's a third party and somebody could theoretically get at it. I mean, I think Nikki, you shared with us in the pre-show, uh, LinkedIn post where somebody kind of flagged that issue with like, if that doctor is taken to its logical conclusion, everything you have saved on a cloud is theoretically no longer privileged and that can't possibly be what we do.
>> Well, and it makes me think of a, um, there is a case from two that's ethics opinion from 2008. I understand ethics opinions, they never talk about privilege. They always are like, "Well, we can't address privilege because that's not an ethical issue." But the underlying rationale is analogous. And this was an opinion about Gmail and a free Gmail account and the ads that are served up to, uh, based upon the content of the email. And this was in 2008. And the, um, issue was whether it was violated confidentiality to use free Gmail to talk about confidential information, uh, with clients. And they said that since it's only a machine scanning it to serve this information up, not a third party, that it doesn't violate confidentiality. And I understand there wasn't privilege, but I think that the same rationale applies here. You know, it especially, you know, the, there's a line between what's public facing, what's not, and there's all these different tiers that are available on the public tools, the general use tools, um, in terms of pricing tiers and in terms of the different settings you can put on the back end. So I, I don't think that that is a black line rule you can use anymore. Is that the expression? You know, you can't draw.
>> Yeah, you can't draw a line there anymore.
>> Line the butter.
>> Bright line. Bright line. Yeah, that's it. Um, but you know, that's I, I think that and it again, the cloud, I think I agree 100%. Like no matter what tool you're using at this point, in theory, third parties have access to it on the back end, depending on the situation and depending on the contract and the terms of service. And so I think it's just kind of it certainly isn't going to withstand the test of time and I don't even I'm not even sure that I agree that it applies with the state of things now.
>> Because doesn't that then open up the possibility any any any statement you using technology, you then have to start to interrogate who had access, you know, did was it was it confidential, which is not enforceable, you know, long time. It's just not practical.
>> Yeah. I would just kind of kind of underscore what what both Nikki and Joe said, I think, which is like, doesn't, you know, kind of what if you reversed it? What if the lawyer had used Thompson Reuters's Co-Counsel to write a client strategy memo, uh, containing privileged information and then sent it to the client, um, and then somehow that, you know, the other party got access to it or or something else. I mean, why would that be, would would it, does does providing privileged information by anybody, whether it's a lawyer or the client, to an AI tool immediately break the privilege under Raikoff's ruling? It seems like it would.
>> Well, think also from the other standpoint, like what if what if they just took whatever Claude gave them, put it back into a Word doc, and said, "Hey, no, I generated this myself." And does that all of a sudden protect it?
>> Yeah. Like like had had he created completely by himself these documents, it would fall under like there's a case that's cited in the brief. I can't remember what it is that talks about how the client's notes in anticipation of a meeting with a meeting with and communication with a lawyer are protected. So yeah, had he written it in by himself, he'd be fine. And that's yeah, that's what makes it.
>> There's like multiple different things going on here. There's like the the the notes and then there's also whether or not putting things into AI is a different breach of privilege right there. Right.
>> Right.
>> But I think it also depends. I mean, if are you putting it in in ChatGPT? Are you putting it in a different? I mean, we've been telling people from the start, don't put privileged information in ChatGPT. Why have we been telling them that? Because it might get out there. We've been saying that from the beginning. So that sort of I that to me follows this exactly. We've been saying it might get out there and now we're saying, yeah, it might get out there. That means it's a potential third-party exposure to this. Um, yeah, also in this case, I don't know the facts of this particular one enough, but I think Joshua mentioned it in the comments too, like it didn't sound the way this is written up, it didn't sound like the guy knew he was having a meeting with a lawyer. The lawyer like bring your thoughts. He did it. He like he just sort of the lawyer even said he did this thing on his own and then he came to me and we had this conversation and tried to claim it was privileged. So depending on I mean, we don't have enough facts in there to know exactly like if this guy was like, oh, I might be in trouble, let me look up all this legal stuff and then I suddenly have a lawyer and then I gave it to the lawyer. Like that doesn't work for me.
>> So so he so he knew he was the target of an investigation and he had already commissioned, uh, Quint Emanuel at this point. Uh, but the re the documents he created for the purposes of their meeting, he wasn't told, hey, go out and create a bunch of documents, but he did create a bunch of documents that he brought to the meeting was like, here's what was going on, like to explain it to them. Uh, which yeah, which I I feel like that's the sort of thing as a public policy, like as a somebody who used to defend people, that's the sort of stuff that you kind of like when a client comes in and has a sense of what happened and can just walk you through simply like what their story was. Uh, and so you don't want to let necessarily frustrate that if you can avoid it.
>> Well, and you also want, I mean, how can you decide who uses what tools to prepare documents at this point or to bring put their thoughts together? I mean, everyone's using generative AI tools to help them flesh out concepts or to help them make things look pretty faster. Do you know what I mean? Rather than typing it out themselves. And I, I just don't see how you can, um, and you can use ChatGPT depending on how it's set up. If it's an enterprise tool, then it's absolutely no different than using a legal, you know, specific tool. So I, I think it's just that's one of it just reminds me of so many of these different, again, I'm thinking more ethics opinions rather than, um, on the legal side of things, but it's not going to withstand the test of time. There's no way it's not going to withstand the next six months. I mean, it just seems to me like it's it's silly to to reach these very specific conclusions rather than sort of more general propositions. And I also think, like we said, the public, you shouldn't discourage people from trying to, um, communicate more clearly with their lawyers. And then you're going to have to start talking to your clients ahead of time at engagement and put it in the engagement agreement about don't ever use generative AI to collect your thoughts, you know, before you reach out to me. That just seems to go against public policy for sure.
>> Or as an example we've used before, what if he used Grammarly to correct it before he sent it off or something?
>> It's just going to be going to be pervasive, isn't it? It's going to be everything used by everyone. And it's going to be impossible to work out.
>> Where it is. You know, it's just not enforceable at all. Yeah.
>> Also leads to the question on, um,
>> Like, you know, like
>> if a judge understands the difference between, um, a closed GenAI system and an open GenAI system too, and the tiers and like a Cha or a Claude, um, and just, you know, based on the judicial disclosure rules, um, around, you know, like when attorneys use AI to draft stuff that they present to the courts around the country. Like I get the sense that a judge really doesn't have a strong understanding of this at all. Um, and to Joe's point too, like how is this any change from a Google search as well? I mean, all the search engines are being quickly replaced by chatbots. I mean, like I don't even do Google search these days. I just go on ChatGPT because it's so much easier. And like Google search too also has, you know, Gemini all the time. So it's everywhere and it's, you know, it's completely integrated into everything. So like I still see like the industry being like, oh, it's this whole new tech or whatever. Um, and it's already 2026 and it's in everything. Um, like it is akin to Google search was in like the '90s.
>> And it's so but I mean, what I've been saying like up till now is that I mean, it is whether or not you you ultimately this decision stands, I feel like all along people have been saying, well, that's crazy, no one would ever think that, of course this stuff is privileged, of course it is. And I've been the one saying, you need to be aware of this because this might come back to burn you. And now it has because it is creating this technology is different in that it is creating all new data and outputs that other technology was not doing. So, it was going to be a new thing judges had to figure out how to deal with. I mean, the analogies work, but they're not the same. People have been creating this data all along just assuming it was all going to be safe and privileged. And I've just been like, no, don't put things into ChatGPT if you really, really don't ever want them to be seen. And so it's like, to me, whether or not the ruling stands is interesting, but like the fact that people are shocked that this could possibly have been an outcome is what really gets me because some of I mean, we've been talking about on the show for a long time.
>> Yeah. One thing I thought was interesting, and maybe I just didn't read the story carefully enough, was that it seemed that neither party raised this issue, the issue of just sharing it with with the LLM in their arguments for or against why the document should be privileged. It it sounded as I read it, neither party had brought that up and Reikov kind of brought that up. Suante, is that the right word? You know, was that was that raised?
>> No, that the the government's brief does, uh, talk about how he used Claude. Apparently though, but you're right that it took a while to get to this. Apparently, this dispute was going on for a long time where the defense was claiming privilege and the government was like, "Okay." And they had it sequestered and hadn't looked at it yet. And then eventually they went to the court and were just like, you know, we've been fighting about this, can't reach a resolution. The crux of our problem is this AI question. What do you think? Uh, and that's what he answered.
>> Yeah. All right.
>> All right. Clearly, sorry. In the privilege log, very clearly identified it as artificial intelligence generated analysis conveying facts to council for the purpose of obtaining legal advice. So I mean, they like put it in the, which I think is, I mean, I, I applaud that. I applaud the honesty in doing that. But I mean, if this becomes a thing, you have to start wondering if people are going to stop putting that specific of descriptions in their prologs.
>> All right. Well, that was a story about a, uh, a client who, I guess, was was not lucky from his use of AI. Um, Reese, you had a last minute entrance here in our lineup of stories for today in which maybe a judge was not so lucky from his use of AI. You want to talk about that?
>> Yeah, sure. So, um, Judge Xavier Rodriguez, I believe from Texas, um, he, um, was interviewed in the Wall Street Journal in January, and he was talking about how he tends to use AI, and it's really just for, you know, like I saw as akin again to like a Google search, just trying to, you know, like find some info that's online anyway. Um, you know, I'm always going to keep going back to this. You know, it's crazy how people don't freak out about doing a Google search about something as opposed to asking a ChatGPT or a Claude. Um, and the Fifth Circuit expressed their concerns about him using the tool and they were like, you know, these AI tools would never replace, you know, Jud, um, the expertise of judges and all that. Um, and it just seemed like an overreaction. I mean, you know, just going through all our stories this week, too, it seems out the problem with the industry as a whole is that they don't read the articles we write or like check out this show, cuz the reactions to people just, you know, you know, super innocuously like going on a chatbot or a Claude to, you know, to ask questions which one can ask anywhere else on the internet, you is just a very emotional reaction and a complete misunderstanding of how these tools can be used too. Um, and it is disheartening, I guess, to see the lack of education on these tools even in 2026. Um, and there's, you know, like we saw this too, like last week when Claude released its login and everybody freaked out and like all those legal tech stocks crashed and then everybody was like, oh no, that it's it's actually okay. It was a complete overreaction too. There's all these weird emotional freakouts over AI tools. Um, and I kind of read the article as the Fifth Circuit is just concerned about what, you know, Xavier Rodriguez was doing with, um, this tool and he was very clear what he was doing and like how he was doing it too. And, you know, you've seen Judge Rodriguez talk at a ton of, um, conferences, too. Here's one person I would be confident with, you know, in terms of trying to understand how to properly, um, you know, query an AI tool. Um, I think the problem is that there is no structured, you know, guidance on these tools and there's no like standardized educational either. Um, a few weeks ago, actually, I think it was, um, Judge Agel, and I'm totally destroying that name.
>> Yeah.
>> Yeah. Um, from Louisiana and a few other, um, just too from the state courts and the federal, um, side as well, who, uh, came up with this like judicial group to kind of standardize how to approach AI. Um, and, you know, those are the types of efforts that I feel the industry really needs because it's kind of just laying around. Um, it's kind of like we, you know, we threw the industry into the deep end and they're freaking out, but the deep end's really only three feet and everybody can just end up and be okay.
>> I mean, listen, good news. This was an article that I was I submitted a different article over this week, but I just put it in the chat. New York, um, issued, uh, 154 pages of guidance for judges, um, uh, earlier, I think it was, um, in January. It's the judicial guidance, um, was issued for the New York State Unified Court system and it's the interim policy. So, I mean, listen, if we're going to issue guidance, why not 154 pages?
>> You you you can use you can use ChatGPT to summarize that.
>> It feels like they're just hanging on. It feels like both this and a little bit the last story feels like, you know, it's just people are going to use it. Judges are going to use it. They're going to in the same way as they going to use Google like we've talked about. I understand that there's a difference and but it feels like, you know, there's got to be a recognition that that in the same way as everybody, the whole world is going to have to move on in the way that we do things and what the workforce is going to move on, judges are going to move on, they're going to be using this and okay, fine, maybe there needs to be some guidance in terms of not handing over to AI, but equally, I feel like my my most sort of, um, extreme view would be that actually they realize that that a lot of the work, the very special work that lawyers do. This could be really deeply unpopular thing to say is actually going to become available through AI, right? Like that actually, you know, like I know that there's going to be a question about data and accuracy, but actually, you know, a lot of the stuff that lawyers do to a great extent is going to be available and maybe, you know, pushing it to the extreme, judges do. I know that that is within the construct that we're talking about is not the case, but do you know what I mean? I feel like there's this like clinging on and they need to move with the times and they need to accept that actually even judges need to think about staying relevant.
>> There's just no go ahead.
>> I I just real quick, I I read this more as just slander against Judge Rodriguez for his decision. Uh, Edith Edith Jones, that that panel was entirely about, you should allow Republicans to cheat in elections, was the decision, and she was mad that a Republican, Xavier Rodriguez, mind you, had ruled the way he did. And the whole thing was just a series of character assassinations on him for making his ruling. And I thought this was just part of it.
>> Yeah. You you know, the whole idea that somehow that the the judge is is, you know, handing over his decision-making to AI is is just ludicrous. I mean, it's it's again, it's really no different than when a judge has a law clerk write a draft of an opinion. Uh, you know, or or uses any other tool. They're still going to review it and and sign off on it. I mean, this is like the the AI arbit to go to the AI arbitrators that you know, Bridget Mary McCormack has has introduced at the AAA, you know, there's a a heavily AI process involved in producing these arbitration awards, but then they have human arbitrators read them and review them and sign and make any changes or whatever else before they get issued. And I, I just don't see what's wrong with that. I don't see that the AI process is really all that much different than using a a law clerk in the process.
>> Well, it's almost like they were looking for they were it's almost like they were looking for a reason, like because I mean, because ultimately, look, Grassley's statement was pretty benign, right? Judges should have policies, uh, governing AI. The federal court should have a policy, maybe, you know, I mean, it's just or or at least have guidelines for to make sure it's not being improperly used. I mean, that's that's that's a reasonable, um, that's a reasonable statement, but then kind of taking it to like, kind of what what they were using it for, kind of like a codel to like to like attack this judge for like you said, just doing just just, you know, just using using a different way of like draft of like researching an issue. It's not like it sounded like he was like, hey, ChatGPT, decide this case for me. I mean, then okay, yeah, I could I could see why people would be mad about that, but that wasn't the case.
>> Yep. Uh, all right. Um, I think the next group of, uh, unlucky, uh, unlucky, uh, not beneficiaries, what's the other side of word of that of of AI are the, uh, folks, maybe the folks at Baker McKenzie who, uh, were part of a massive layoff there this this this week that the firm is at least attributing to AI. But Joe, you have your you're not so sure about that.
>> Yeah. So I I mean, I think I think to be fair, I I think that the firm did use AI. I I do think they thought it was had something to do with AI and I think they understood that PR-wise it's good to use AI. Um, but my take on it was they, so they laid off all these staff members across the the globe. Baker McKenzie, very large firm. Uh, and that's what what really got me thinking is as this moves forward and as people increasingly either want to lay people off because of AI or want to blame their otherwise their otherwise layoff plans on AI. Uh, is AI going to impact firms differently? And I thought this was an interesting test case because there are two kinds of firms at the very top of the AmLaw 100, right? There are well AmLaw 25 at this point. Uh, there are the firms that are making tons of money, uh, with a couple hundred lawyers, a few hundred lawyers, and they're billing like crazy to do the work that you can't do without. And there are firms in the top 25 who make revenue hand over fist because they have 5,000 lawyers in 130 offices around the world. Baker McKenzie being in the the latter, uh, category. And it struck me that I, if AI is capable of not replacing lawyers, because I don't think it can ever do that, but, um, creating efficiencies so good that fewer lawyers are needed to get the same amount of revenue, uh, that is going to impact firms in the Baker McKenzie, DA Piper, Denton sort of, uh, business model way more than it's going to affect the Creasaths of the world. Uh, and the reason I think that's true is that they're the you need to start looking at what the RPL is on these firms and like what what the revenue per lawyer is because as things get more efficient, uh, if you're making 4 million per lawyer, that's probably not a factor of grunt work that you're making that money. Uh, but it probably is a factor of throwing bodies at the problem when you're scraping by making a hu billions and billions off of 700 grand per lawyer. Uh, and so I think we should start thinking about whether or not it's going to impact these firms differently. We've kind of had a a vague like maybe we've talked about big law versus small law, but we haven't really talked about even within the top of the AmLaw how it's going to impact folks differently. And I think the reason this is really relevant and important for people to think about is we're in the midst of a couple years worth of watching mega mergers go like crazy where these like firms in the lower half of the AmLaw 100 are getting together and popping themselves up into the AmLaw 30 or so, uh, by just sheer force of bodies. The idea that that model is going to be successful long term is certainly complicated by the idea that AI can start to chip away at whether or not bodies are critical.
>> So the formula is the lower the RPL, the higher the likelihood of replacement by AI or
>> I think that's probably true. Yeah, I think that's probably true because I, I think you you have more risk in that kind of model than you do in a model where, I mean, if WCTEL gets a little bit more efficient, uh, they're just going to pass they're just going to like pass it all on. They're going to keep they're going to keep getting the money they're getting. Uh, cuz lawyers like money, uh, and sorry, spoiler alert. Uh, and so when the they get more efficient, they're just going to find a way to collect the same amount of money. That all works. Uh, if you're already in a tier where people are paying a super premium for your lawyers, uh, if they're in you're in a tier where you're a lower price point, but you're getting big revenue by bodies, then that means you're already at a place where people are a little bit worried about whether or not they're going to pay extra for that. And if the AI does it faster, then maybe they don't need it. Makes sense.
>> Yeah. I And I will say I just, to throw in my other story real quick, I, the other story I potentially was going to talk about was kind of a flip side talking about a boutique firm that, you know, didn't replace a lawyer with AI, but had a lawyer leave a six-lawyer firm. They they lost a lawyer and rather than hire a new one, which for a small firm is an expense, right? Like to find a new person, train them, like it's a risk also, uh, they just said, kind of, do we think AI can take enough off of everyone else's plate basically that we don't need this extra whole body, uh, and it turned out to be a success story. And, you know, it's not they're not replacing somebody, but, you know, they got got a little bit better and they saved some money and continued actually ended more profitable at the end. Uh, and that's the other kind of model where maybe the smaller you are, the more likely it is that AI is not a threat to you, but a leverage as opposed to the big people.
>> Well, listening to you talk about these and describe what's happening, I know we've talked about this before, but I don't understand how our profession in general is supposed to keep up with us in terms of the law school's pivoting in time and, um, figuring out what does a career look like for a recent grad and how are they supposed to get the experience and how do they get the training and how do and how do larger and even smaller firms when they need to have experienced people, where are they going to get them from? Like especially in this gap where these people are trying to gain experience apparently by going solo. I don't know how they're going to get the experience, but what is it, what does that look like in the next five or six years? I can't imagine what our profession is going to.
>> It's not going to be able to pivot quick enough. There's no way. And I don't know what that looks like from a practical standpoint, you know.
>> See, I was going to go to your story next, but now you just queued up Victor's story, so I feel like we have to go there. Listen, maybe we can, you know, maybe we can just replace you and we'll all cue the stories up. I'm kidding.
>> I'm I could easily be replaced by, uh,
>> I'm kidding, but I'm happy to queue up the story and go.
>> Victor, that's you.
>> I was thinking about the Verine model thing. That was interesting because I never I never thought of it like that. Um, hm, cuz cuz sorry, just to just to kind of stay on that topic a little bit. I mean, but like isn't one of the reasons for Verine models cuz you want to, you know, you want to have like that big global, um, footprint, but you don't necessarily want to have like, you don't want to deal with like the hassles of like with all the different regulatory schemes, all the, you know, maybe the different profit sharings, that kind of stuff. Um, I mean, would that necessarily be affected by? I don't know. Yeah, I, because it seems like it seems like it's almost like a regulatory thing, not necessarily not necessarily a quality thing, but I don't know.
>> I think the Verine part of it definitely is regulatory as opposed to I, because not all of I don't think Baker's not, I think, but the others are. I wasn't one of those. Yeah, one of one of those is not and the others are and obviously the Ver Oh, Baker is okay. So, one of one of the big ones, I can't remember which, uh, is not, maybe DLA, I don't know. Anyway, one which was in the news for the wrong reasons this week, um, but, uh, one of one of them does it. But yeah, that's a very regulatory thing. But but the just the si the size of it, I think, is is the issue, cuz you could have an international firm with lots of like sub do that Verine thing where they're all like separate companies, but not sort of, uh, but not necessarily be 7,000 lawyers or whatever it is.
>> But yeah.
>> Yeah. Sorry. I, I was interested in that part. Um, no, it is, you said as Voltron. I was like, "Oh, that's a good, that's a good, that's a good analogy."
>> Yeah. Right.
>> Um, but anyway, yeah, the story that I I put in there, um, talking about a Bloomberg, a Bloomberg Law survey that talks about, um, you know, artificial intelligence and what is expected from people graduating from law school. And this says 76% of attorneys expect law school students, graduates, sorry, to be able to cite AI generated materials. 63% expect an understanding of AI related legal ethics. Only 14% accept skills related to prompt engineering, which is kind of, you would think that, you know, you'd want them to be able to understand how to actually, um, how to actually, uh, you know, set up a prompt so that it's responsive, but, you know, it just shows again, like new lawyers are are expected to be AI savvy, lawyers are expecting it now. It's become a much more, you know, I mean, you know, we see these surveys all the time, but it's always good to kind of put numbers on things and kind of and kind of, you know, um, um, yeah, kind of let people, you know, let us know sort of like where, you know, what what people are thinking and how how things are looking and whatnot. And this definitely, you know, um, reinforces the fact that, you know, that that that young lawyers now, especially graduates, are are expected to know this stuff and that it's not going to some be something that they can just avoid. Um, yeah, the the prompt thing kind of surprised me a little bit, but, um, you know, that could also be just people just not really understanding, um, the importance of it. Um, but everything else was was pretty, you know, seemed like it was it was pretty spot-on.
>> I think I think it's interesting just because it really is only a couple of years ago that we were probably on this show talking about the fact that at many law schools, they were debating whether to whether to teach AI, whether to let their students use AI in any way. Um, you know, whether to penalize students if they discovered that they were using AI. Uh, and now it's almost like, you know, the the profession's come full circle in terms of recognizing the critical importance of teaching these skills. You know, somebody asked in the chat, are law schools doing this? I mean, I think that's really an issue. Some some clearly are and some are clearly doing it, uh, extensively, but I, I my understanding is there's still a lot of schools that really aren't doing much in terms of teaching these skills yet. So, you know, as as with technology before GenAI came along, a lot of law schools were behind in teaching those skills. Uh, it seems that there are a lot of them, I would say, I would venture a guess that the majority of them are still behind in teaching these skills. This makes sense to me in terms of like the so the site when they do their own research, if they, you know, presumably they a lot of them are being taught how to use AI and there the site checking within the context of a law student is going to be critically important to understand that if they're using AI generated material, and then the understanding of ethics, well, I think that's probably underpins everything, doesn't it? You know, you've got to at least come into the profession with an understanding of all the ethics is. And then but and then prompt engineering, to me, seems more like practical, you know, when you get into the workforce and then you're actually doing the more practical. So I don't know, law school is there is an element of practical, but like I, it doesn't surprise me perhaps as much that you're you're not, you know, you're not sort of putting it into practice in the way that you have to then when you join a a law firm. I don't know if anyone disagrees with that.
>> Um, the I feel like, you know, CoCounsel and the Protege tools and all those research tools, they have been, you know, changing so that prompt engineering skills aren't as required because the tool kind of like, you know, like has this interface where you're able to talk to it in a very simple way and it can, you know, kind of take that stuff and basically create its own prompt too. So like there has been a growing effort to like, you know, to, um, to create tools that are just like super easy to interact with too, um, and to kind of decrease the need for prompt engineering skills. I will say that these days, there's still a huge need for prompt engineering. Like I'm on, uh, Copilot and Copilot is the stupidest goddamn thing I've ever seen in my life. Like it just take, you know, like I have to be very, you know, specific, I guess, with like everything I say and it takes ages to really write those prompts too. So I do see a need for some prompt engineering skills just cuz quite honestly, I find some of the AI tools just kind of stupid. I mean, like for like all that cash and I'm like, this cannot be it.
>> Um,
>> Yeah. Um, I do feel that there is a change though that's hopefully coming soon.
>> Yeah, I mean, that's kind of my experience with Claude Co-work. It's just like it's supposed to replace prompt engineering where you can just kind of say the thing and then it figures out the right prompts for it. And, you know, like all the studies suggest as a normal agentic thing, it gets it right till like task two and then goes wildly off the rails. Uh, and you end up having to kind of go back anyway. So, for now, you still need to prompt engineer.
>> Yeah. All right. So, we talked earlier about, uh, you know, a judge who was called out for outsourcing responsibility to AI. Nikki, you've got a story, a a warning shot out of California. I better duck because I'm out here right now on whether lawyers can outsource responsibility to AI.
>> Yeah, it's I I thought it was interesting that California's, um, Senate passed some, uh, legislation that sort of codified what lawyers should and shouldn't be doing with AI and it's been sent over to the assembly to review. But, and it also talks about arbitration, which also caught my eye, like the way that arbitrators should use it. But, you know, it, it, I didn't really touch upon this so much in my article. I just sort of covered what what it was, uh, the language that it was using and the requirements it was imposing. But, you know, it's a little strange to me that it's I just feel like the industry is having these knee-jerk reactions like this is out of control. We got to deal with this somehow. And so, you know, there's ethics opinions now, they're trying to, you know, codify it and create some laws and maybe that'll have some impact and make the lawyers pay attention. It, it just seems like a little bit of overkill to feel like you need to pass a law. And I, I'm not quite sure what, I don't think they talked about. I wonder what happens. I don't know if it didn't occur to me just now. Like what happens if you violate the law? I don't think it's a criminal law. I, it's, I actually am, I don't know why this didn't occur to me when I was writing about it, but what kind of law are they talking about? Is it civil procedure? Like I don't know.
>> Ask ask opening GPT, I don't know. I think I
felt like we do. Do we talk about? Was there another state? I feel like we either talked about this before, or maybe some other state was also doing something like this. But I mean, yeah, I agree. It just doesn't seem like something where you want to legislate, uh, in any way. You know, let there be ethics opinions, and you know, some of them will be wrong, and some of them will be right, I suppose. But it just doesn't seem at all an area where you want to legislate, especially, you know, we just talked about another context, but whenever you've got any kind of a technology that's rapidly evolving and rapidly changing, the last thing you want to do is lock in a bunch of rules in a statute that then has to go through a whole legislative process to change.
It just seems like overkill. I feel like the problem's becoming so bad and it's so evident because there's a new headline about it every day that there are just these knee-jerk responses that try to rein this in somehow or try to get the word out. And clearly, it's not clear, like someone referenced, really no one's listening to us, it's not getting out there. But there are some really, there are some really interesting studies, aren't there about not just lawyers, but how people are becoming already. Well, we know that people are really reliant, but like it's actually changing the, you know, your brain in terms of, um, you know, becoming really lazy. So that people are actually, and lawyers are only human. I know we're, you know, amazing, but people are just becoming, it's lulling your brain almost into this, lulling your brain into this false sense of security, you know? And it's kind of bizarre to think we're only at the foothills, you know? And already like people are doing. Initially, it was because we didn't understand how it worked, you know, and there was all of the, oh, we didn't know it could listen. And then, and now I think you're almost at the stage where some people are just becoming really super lazy, you know? Like I think there's obviously differences, but already you're seeing people going, "Oh, yeah, why would I do that when I can do that?" It's, it's really a problem.
I have to say that I was, well, before AI, so it hasn't changed anything. So it only impacts so much.
You're ahead of your time, always.
I like that. Um, so last week we talked about Claude maybe encroaching a little bit on the legal tech world. Caroline, you've got a story this week of maybe some signs of OpenAI doing moving in that direction also.
Yes. And this hasn't actually been written about. Like, I know it's funny because with when Claude, um, when Anthropic Claude's legal legal plugin crashed the market, everyone was like, this is just so predictable. This isn't really a story. Why is this, you know, why is this a thing? Um, which is, which is true. But, you know, it is also a thing, right? Like, so a big legal tech company entering the, sorry, a big tech company entering legal, albeit whether we think it's good or whatever. But it's a big thing. Anyway, so now OpenAI, I've heard from somebody, um, in legal, in the legal in-house legal world, who, um, has been contacted by OpenAI. I was a bit fudgy around the edges just because it's all off the record and just like that initial. I thought I thought it would get more reaction. I'm not going to lie. It was a bit crickets on LinkedIn. Perhaps I should have made the headline a bit more spicy. But I was trying to be sensible, you know, because obviously there was so much backlash against Claude, and we realize that this is a bit stupid, you know. Um, so I was trying not to be sensationalist, but it's interesting nonetheless because OpenAI is reaching out to people in the legal sector, particularly in-house, asking them to test. Um, what it is is there, so they also call them skills. So, um, Claude has skills, which, and but this is not as, what OpenAI is doing is not as evolved yet, but they're obviously going down the same trajectory. So they're creating task bundles. They're asking people in legal to test, um, and, um, work with them to build these out. It's called, it's, um, the communication, they call it, um, it's at the alpha stage. It's pre-beta, so it's really early stage. Um, but they're asking them to join the early testing program to build ultimately these legal workflows and agents. And in some ways, you might think, because we didn't really know what's happening with OpenAI apart from we know they're doing something. So it's kind of interesting to see where, what stage they're at, which is still very early doors, but it's also interesting that there's absolutely no doubt that they're going down that path, um, and that they're very much sort of now reaching out to people in the legal sector to help them do that. It's also interesting. I mean, we, I, you know, go back back to the prehistoric days of generative AI. I mean, OpenAI was the organization that originally provided the funding and the resources for Harvey and for Co-counsel. And I mean, they, they kind of put, you know, went into legal sort of in a big way in those early days, or they seem to see potential of it, and they kind of, it feels like they kind of then turned it over to them, you know, rather than try and build on that within the OpenAI company, or maybe they saw better, more greater riches in other directions. I don't know. But it would make sense for them to do that, I would think.
I have a sense that like OpenAI has just been like confused about how it's going to grow. I mean, like Anthropic and Claude were very focused on coding and trying to, you know, like have like an AI platform for the professional services industry. Um, we didn't really understand where like OpenAI was going because it started as a nonprofit and it kind of just imploded. And after some time, it had all these crazy ideas, including having like an adult like ChatGPT or something, which was horrible and insane. Um, and it seems to be like trying to, you know, obviously follow in Claude's like footsteps too. Um, it's interesting that these like giant AI folks are trying to enter the industry. I mean, like they don't have the expertise. They don't have the data. Um, and we've actually seen some legal tech folks actually integrate with the Claude tool, because they're like, "Oh, it's helpful, except it won't replace what, you know, the stuff that we have because we've been, you know, doing this since the, you know, like 2000s, I guess, and we have all the info, we have all the clients, too. So it's hard to start from scratch." Um, you know, like they do have a ton of cash and they do have, you know, the tech, except that, you know, you do need way more to enter the industry. I feel like you need the expertise, you need the understanding, and you need the data as well.
The people that I'm speaking to are are kind of excited. So, excuse me. I think you're absolutely right. And and what what someone said to me today was something that no one has raised. All your points, Ree, about data, about the platform. This is what Harvey, Lexis, all of the big publishing companies say, right? Particular, well, the different things. So the data will be the publishing companies, the platform will be Lexis and Harvey. But what someone said to me today, because in-house teams are quite often using mainstream Gen AI tools, right? So they're not, a lot of them haven't got budget for legal AI. So this is kind of cool within that context. But what someone said to me today was how much money in terms of time and effort to create, to turn this into something which is really useful, fit for purpose, particularly in terms of, you know, whether it might be security, whether it might be building around it, like in terms of the effort and the getting it verified. And the, I mean, you know, what does that look like in terms of the time taken? And also then, once you've done the build, as with any build, if the model changes, like whereas if you go to a big provider, then the model changes, and then they're responsible for updating it. Whereas if you build something and then the model changes, which, as we know, is changing all the time, then you've got the cost and the time of then updating it. So actually, that, I thought that was a really key point. You know, like the in-house teams, I think are quite excited about not having a big legal AI cost. I think the cost is a huge factor. But actually, how does the cost really compare when you think about all the work, extra work? I think that's kind of fascinating.
Well, I think the cost is going to be the thing long term here because I don't know if people have been following the, like somebody did some analysis of what it would take for this technology to pay for what it's both, what it currently has and the capex that it's projected. And it's something like they're currently absorbing losses because they're trying to build market share. And somebody just like worked out like to pay their current and on the books for the future costs, what would the subscription have to be? And that's the thing. Like these folks using these can, you know, wanting to use the actual founder models. Part of it is that they don't have to rebuild everything every time, but also part of it is it's slightly cheaper than the bells and whistles being built around it. But what even does that cost ultimately? And somebody worked out that it, it works out that basically your $20 subscription probably should be about $1,100 a month to cover the costs to make that cut work out. I don't think a lot of us are going to pay $1,100 a month when that happens. I think if you asked Dario and Sam Altman and all that, they would say of course everyone will because it will be doing everything. But given that it doesn't and it can't, that's probably not going to work. But for the people who are paying for it at an enterprise level, like they're getting a cheaper deal now than the bells and whistles being built around it by real bespoke legal folks. But those days are limited too, I think. And they're going to, they're going to squeeze them before they squeeze the consumer people, you know, because that's how it works.
Does that factor in the potential ad revenue that they're all banking on?
Well, only ChatGPT is banking on that so far. But yeah, so far.
I mean, you know, we don't pay for Google search, right? And that makes a crap ton of money. So.
I mean, Claude made all those Super Bowl ads.
Right. All right. Well.
That was my biggest disappointment at the Super Bowl. Even, even worse than the Patriots losing, was that they only showed like the 30-second snippet of the Claude ad instead of the full ad, which was, I thought was really funny.
And apparently it backfired. It was the least popular ad of the Super Bowl.
Nobody got it.
Well, because apparently we all tech people thought it was hilarious, but the public was like, "Why are you making fun of ChatGPT?"
Yeah. Yeah. It didn't play well. Well, what it makes me think about though, is if we all hearken back to 1995 and think about all of these different, um, public facing free tools, things that we use in our day-to-day lives that are available that make money, a lot of money for these companies. They're all monetized in ways that we never would have thought, right? I mean, monetized just by the fact you become the product and they're selling your data, or it's the advertising. And so I, I would not put it past these companies to somehow perhaps use their own tools to figure out a way to monetize it, you know, in a way that we otherwise would not have envisioned. So, um, I mean, we'll see. But it's, you know, selling your privileged communications to the government. That's one good way.
All right. Any any final, uh, just down to the wire here. Any any other big news we didn't talked about or any comments or anything else anybody wants to say before we bring it to an end?
Um, Harvey.
On the non-AI front, um, you know.
Oh yeah. I put this on as an alternative thing. The social media trial going on in LA. Apparently being on Instagram 16 hours a day is not an addiction. I, I, I, I was, I was quite, I can't believe someone actually got on a witness stand and said that with a straight face.
I agree. I'm only on that 12 hours a day. So I'm good. I'm good. Yeah. And and Ree, if you're going to bring up Harvey raising more money, that's getting to be like the dog bites man story. I mean, who even cares anymore? It's like.
Just throwing cash at it.
Yeah. All right.
I will say though, and I have to clarify this, um, the investor in this round seems to be, um, tied to the Singaporean government. So, Singapore potentially could have invested in Harvey. I just need to confirm. I, um, trying to reach out to the fund and ask who they are exactly because there could be a few people. But, you know, that could be interesting if, like, the, um, the capital fund of Singapore, um, invested in Harvey. That's pretty huge. Yeah, they've got that, they've already got that relationship with the courts in Singapore, uh, for kind of small resolving small claims courts issues or something. Um, so yeah, that'd be interesting.
Which all end in canings, I'm assuming.
Does anybody remember that? Like, am I the only one old enough to remember that?
That was a big deal.
That was a huge deal at the time. You're a lot older than I am, Joe. I don't know. That was before my time. Yeah.
All right. Well, thanks everybody and, uh, we will be back next week to, uh, talk about more Harvey investments and whatever else happens in the world of AI. See you all then.
Thanks everyone.