Takeaways from ILTACON, the 'Hugging Face Hack' and more

Speaker A: Welcome everybody to Legal Tech Week for September 4, 2020, 26. 2026. Um, not 20. 2026. That would be way off in the future. Right? This is Bob Ambrogi. And uh, this is the show where we talk about top, top stories in legal tech and innovation for the past week or so. And I, uh, think we're going to be talking a little bit more about uh, Ilticon, uh, in this show, uh, where uh, most of us were. Victor, uh, sorry you weren't there, but uh, the others of us were there and I know we've all had thoughts on it and uh, some other stories as well that have come up. Uh, and uh, by way of, uh, introducing the panel, let's get started. Nikki, why don't you kick it off.

Speaker B: Uh, my name is Nikki Black. I'm the principal Legal Insight strategist at uh, uh, 8:00am, the team behind my K Law, my case law, pay case here and docket wise. And I write legal tech columns for the ABA Journal, above the Law and the Daily Record. And I am looking forward to a three day weekend.

Speaker A: Two day, three day Mondays. Did you say three?

Speaker B: Three day? Yeah, three day.

Speaker A: We're all a little tiny more, but

Speaker B: I can count that.

Speaker A: Okay, good. I can't hear very well. I guess that's my problem. All right. And Victor.

Speaker C: Hi everyone. Uh, my name is Victor Lee. I am assistant managing editor for the ABA Journal. Uh, looks like I missed. I missed, I missed a great party. Uh, kind of disappointed I didn't get to see show crow like you guys did, but looks like, you know, I guess we'll talk about that.

Speaker D: We will.

Speaker E: And um, Joe, Joe, Patrice from uh, above the Law and the, the uh, podcast. Thinking like a lawyer. And I am, you know, trying to recover from Ilta, though it's hard to. It's hard to even get back into normalcy after that. But we're going to do what we can.

Speaker A: All right. And last but not least, Steve.

Speaker D: Hey, Steve Embry. I write for above the Wall and I also write a blog. Take well Crossroads and like the rest of you, Mark, except for Victor, I am slowly digging out from the week of fun and games at Ulta.

Speaker A: Deltacon was a lot this year. It felt like it was, it was really intense. Uh, and I had to be there a few days earlier. Even so it just got to be, uh, by the time I left, I was very glad to leave Nashville. But um, uh, we actually, some of us did a panel live on stage at Ilticon on Thursday, the last day of the conference. Uh, which I posted the video of. Not, uh, the greatest videography job, uh, in the world, but, uh, get the point and you get to hear us anyway, talk about, uh, some of what we saw and heard and thought about the show. Uh, and I think we're going to talk about it a little bit more today. Uh, for anybody who's wondering what the heck this is behind me, who was, if you were at Altacon, you know darn well what it was if you weren't there. Uh, this is a view of the Harvey booth, which we'll probably talk about a little bit here. But both Harvey and Lagora had these kind of over the top booths, uh, with Harvey stretching on for, uh, what did you say, half the exhibit hall Joe, or something like that, uh, taking up space for sure. And I think both Joe and I, in our posts about Ilticon, uh, analogized it to the Berlin Wall. And after somebody read my post, they sent me this photo with the tear, uh, down the wall. Mr. Harvey. Uh, so thanks to Tom O' Connor for sending me that. And in honor of his wife Gail, who he said would have been. Would have been spray painting the wall all week had she been there. Um, so, uh, I don't know. Where should we start? Nikki? Joe, you actually, you guys kind of wrote similar parallel articles in some ways. I think two takeaways were very similar. Nikki, do you want to start talking about it?

Speaker B: Um, well, I really, you know, I always like to look at the big picture and sort of hone in on legal tech from that. And, you know, the big picture in AI is exponential growth. Um, and, uh, I'll have to find the link. I don't have it up. Um, but the big picture, okay, the big picture is, um, exponential growth, right? Like, AI has grown so exponentially. Um, and trying to keep up with the advancements is incredibly difficult for anyone. And then in legal tech, you're seeing that same thing. You're seeing exponential growth. The companies are all, you know, changing course two, three months, you know, every two to three months because it's changing so quickly. And that was one of the themes that came up during the discussions. But what Ilticon really represented for me was this idea that you could see this exponential growth, right? Like the Harvey and Lagora booths, you know, were 10 times the size of any booth we'd ever seen and 10 times the size of their booths last year. Um, the amount of money pouring into the show was astronomical. People were trying to figure out exactly how much each of those booths would cost. Exactly how much it would cost to have Sheryl Cross, you know, sing there like she did. There was a lot of discussion about how much did this actually cost. So there was this exponential amount of money being poured into the show. And then the number of just startups, um, uh, were right at that inflection point before all the consolidation happens. There's just. It's like whack a mole with AI startups, um, and they come and they go, um, you know, they're here and then they're gone, or they're acquired. It's hard to even tell what's happening to some of them. So for me, that was a big part of the show. And then it was just partnerships, partnerships here, partnerships there, partnerships, partnerships everywhere. Everybody was partnering, um, because it's all about data, and no one has all the data. Right. Everyone has, uh, you know, their customers store certain types of data in certain types of systems. And then there's the legal data. And everybody's trying to partner to make it so that their customers can access whatever data they want and still use them, uh, to securely maintain whatever the particular type of data is with all of the specific, um, regulatory and ethical constraints around that data, and also just practical like audit trails or whatever the case may be. So it was just really interesting to see everybody talking about partnerships. Some companies in ways that had never barely partnered before, some of the larger ones are suddenly partnering all over the place as well. So it was, uh, quite a show. And one thing I'll say is that I did not have seeing, um, Sheryl Crow live in a club full of about 200 people on my, um, bingo card for this year or really for the decade. So that was a, uh, really great surprise, especially for someone who was in her 20s, in the 90s. So Cheryl Crow wasn't my favorite, but I liked her. And that was something else. She was a great performer, so I can't complain.

Speaker E: Yeah.

Speaker A: Joe, you had your somewhat similar take on parts of that.

Speaker E: Yeah, yeah, yeah. I mean, similar to the extent that mine used the word polycule. But, yeah, uh, finally we get to

Speaker A: talk about polyamory on this show. I've been waiting for years.

Speaker E: Exactly. Uh, but no, I really did. And that came up, uh, while we were outside of a lunch, I was chatting, uh, and we somehow. I said, this is like a polycule. Uh, and it stuck. Uh, yeah, the nature of this, the jumping, everybody trying to be partners with each other, uh, conversation, which, you know, I feel like even as recent as Legal Week, the conversation was still about exclusivity. Still about owning people's interface. Like we're going to be the way in which you do everything. Uh, and this really felt like everybody abandoned that uh, they, they still want to be uh, maybe the way folks interface but they're not going to try to claim that they own it as much as they're going to try to win hearts and minds. Uh, we will have a platform and we're going to, we're going to show lawyers that they should be choosing us but they aren't going to try to at an enterprise level say we're now the big thing, uh, the only uh, game in town. And yeah, it just seemed like everybody's cooperating. Uh, I compare it to the mcp, uh, the MCP key party. Uh, everybody is just jumping uh, around with each other and uh, it was a different kind of world from where it was even a few months ago.

Speaker A: Was it, you think it was MCP or Anthropic's introduction of MCP and Anthropic getting into the legal space that's kind of precipitated a lot of the sudden desire to uh, get in bed with everybody else or whatever the expression would be.

Speaker E: I think part of it's that, I think part of it is people realizing that uh, exclusivity just wasn't taking. I think part of it is the super hot lateral market. Uh, you all heard this because we were in some meetings together, but there is something to be said for if you've spent millions and millions of dollars to bring over a team and they say oh we prefer some other product, you're going to buy them that other product too because you've already spent all this money on them. Uh, and that just creates a world in which there's not kind of it enterprise level hegemony on most products. I mean I think on the business of law side there is, I think certain, certain functions require that, you know, document management kind of requires that. But ah, for everything else it's uh, I compare it to the Wild west for those of us who are around in the early aughts of Lexus and Westlaw where everybody had both. And it was a battle for which the associates were going to click on and they treated it like pushing drugs. They would get at you while you were in school with free samples. Uh, and you know, hey, you can't beat the, you can't beat the business plan of the drug cartel. Like they're pretty good at it.

Speaker D: Yeah, I think part of it too is you know that I think a Lot of companies really wanted to be the one stop shop and they were working on it, but that takes a lot of time to be able to fill all those. All those roles. And then you come along and you start seeing some of them kind of break from the pack and begin all these partner relationships, which puts them ahead of the pack, ahead of their competitors, who are still trying to push the. Push the pebble up the hill to create everything. And I think a lot of them panicked and said, we got to keep up with what's out there or we're going to left behind. And so I think that was part. I was almost a panic, a panic buy kind of thing. But it was, um. Yeah, it was an abrupt change, that's for sure.

Speaker A: We didn't even talk about the fact that, um, some of the numbers, which, um, they gave us an updated. I don't have that update in front of me, but this was the largest Ilticon by far. Uh, was over 5,700 people in attendance, which.

Speaker B: But I could be wrong right around there.

Speaker A: Yeah, something like that. And then I think Joe had it in his post too. And then, uh, which was like last year was 46, I think, so at least 1100 more than last year in attendees. Uh, the Gaylord Nashville, if you've ever been there, uh, it's gargantuan. And every single hotel room was filled by people attending this conference as well as all the surrounding hotels. And a lot of people were staying in town or at the airport because they couldn't get hotels nearby. Kind of crazy. And the, uh, somebody already mentioned the tonnage thing. Uh, the fact that the exhibit hall, uh, the tonnage of materials brought into the exhibit hall booths and whatever else goes into a booth, uh, exceeded last year's tonnage by three times, even though the number of exhibitors was not three times greater. Only a slightly, Slightly, um, slight increase in exhibitors. But the. It just tells you something about the kind of the scale and the glitziness and the scope of some of these booths there. So it was wild. It was, uh, it was. It was great.

Speaker D: You know, I think. Oh, go ahead, Joe.

Speaker E: Oh, just real quick, I was gonna say, you know, we keep talking about the Harvey and Lagora booths, but real quietly, Fulcrum had as many booths as Lagora. And like they, they just snuck it right on in there. Like, partially because they were disconnected a little bit, but there was a little courtyard in between, but let's not sleep on them. They also had a giant booth.

Speaker A: And also, because nobody knows What Fulcrum does. So that was another thing.

Speaker E: But

Speaker A: no, I had more. I literally had more people ask me that question during the week. What does Fulcrum do? I've never heard of them. Uh, I don't know why that came up so often. But also their booth wasn't so ostentatious. You know, it just, it was. Looked more booth like, uh, even though it took up a lot of space.

Speaker C: I was wondering, because the thing about the MCP that Joe was talking about, I mean, are there any major vendors under that that are kind of resisting that or, um, do you think most people are just kind of embracing it?

Speaker B: There were a couple that were still talking about being in the eDiscovery space, the platform of choice and that were um, very selectively, um, and some of them were only uh, into letting mcps access, uh, admin data is what one of them said, so that people could make administrative changes, but, um, not the substantive data. And it was interesting to hear one company that was opening up and um, talking about how they were just sharing everything with everybody and then another ediscovery acting like who had the exact same type of data in their systems. Oh, you can't possibly do that. It's not gonna be secure. So I mean, how. I don't know. I think it's going to be harder and harder to make that argument if all the other companies are doing it.

Speaker E: Yeah. With this disco in particular was uh, taking the stance that. Which I thought was interesting. They were kind of bucking the trend and their position was like Nikki was saying that there's, there's need to be specific. One of the things they even said was that, you know, no matter how good your MCP piping is, to use an analogy, like, it's just not going to do as good a job at probing the data as their platform will do of probing their own data. Because it's going to come in and try to access it through its intelligence of your data and it's just not going to be as good, was their position. Uh, which certainly, certainly sounds good. And although I honestly thought Everlow, for instance, uh, which is going the other direction. They kind of talked about it as. Yeah, they kind of conceded that, but was also. Yeah, the dedicated eDiscovery professional is going to want to be using our interface, but the senior partner who just needs to check up on things can be doing it through other. And so I don't know, it's, it's an interesting.

Speaker B: I don't know if I agree with their claim Because I think I've talked once before about how I used one of the big legal, I mean legal research is granted it's different types of data and you know, that's just a whole bunch of data that's all protected the same way. Right. There's less of a reason to have certain types of data that are more secure or whatever the case may be. But that being said, you know, when I did legal research using one of the big two and their LLM using Midpage, um, and their LLM and then CLAUDE accessing the pages data, the CLAUDE results were by far better than the other two. So I think that in some ways these um, uh, the OpenAI and Claude, you know, these general purpose LLMs, in some ways they're just more sophisticated in terms, I think in terms of their ability to analyze information sometimes. So uh, we'll see, you know, because there aren't these guardrails and constraints around um, some of what it's able to do within the data but we'll see. But I think that that may not um, stand the test of time that claim but I'm no ediscovery expert either

Speaker A: so, so something we talked about on the panel, we did it there last week was this kind of question of uh. Well in my post I use this phrase we have reached peak legal tech because I had a CEO that I was talking to there used that phrase and I thought it was really descriptive uh, of kind of what it felt like there, uh, certainly compared to past conferences. Um, sort of a funny side note on that is I put in my uh, post that I wasn't going to identify him because I wasn't sure he wanted to be identified. And I have since heard from three people claiming credit for the phrase, all saying it's okay if you identify me as the source of that. Um, but uh, uh, there was only one um, but um, so it all of this kind of begs the question and somebody I think already used the bubble word in the chat here but uh, is this a bubble about to burst? I mean this really was as ah, as extravagant and as huge as we've ever seen on you know, a legal tech conference be I think. And um, where do we go? I mean is Ulticon next year going to be even bigger and even better or ah, are we all going to be uh, wishing we had jobs in other fields by next year?

Speaker D: Are they going to take, are they going to take away the open bar in the media room is really what you're asking, right?

Speaker A: We didn't even Talk about these espresso martinis in the media room. I mean

Speaker C: for those of us that

Speaker D: have covered ILTA before, that was a shock.

Speaker B: Yeah, it was very different. Um, when you're at the peak, there's nowhere else to go but down though. So I don't know. It does. I. There's nothing pretty coming at the uh, on the other side of this I don't think, but time will tell. Um, Joe's been predicting not just legal tech but you know, across the board for a long time now.

Speaker E: So yeah, I, I think the, and I talked about this a little bit in last week's uh, show that the general AI world is in real trouble because it's a series of IOUs being passed around, uh, and that's going to have a real reckoning. And the question is what does that do to, to the downstream legal tech folks who are, uh, we will generously say are a series of rappers. And I don't know, like one of the points I made, uh, I made last week is I felt like this, and in my article this week, uh, I felt like this show there was more quiet recognition of that, that maybe we need to start planning for the possibility that things get dicey and having smaller language models that are proprietary to us, that we control that can keep going if OpenAI goes belly up, or uh, having more agnostic operations or having uh, having data enriching that allows smaller models to get as good a results as big model like that. That whole conversation kept happening and I think it was more about the, the idea that token get in, as I keep calling it, is coming. But related to that is what if everything blows up up top? Which is not too crazy.

Speaker D: Um, it isn't. And I mean the fact that at some point these companies are going to have to make some money presumably means, you know, how are they going to make money? One of the logical places is move away from subscription and move toward consumption and increase your token cost. Now that you've got everybody hooked in, so to speak, let's go to token cost and raise them and that's that. That sort of brings the issue and I think netdocs, Net documents, you know, their, their graph was directed toward this and being able to determine, you know, the, the cost and the, and the benefit ratio particularly, and they didn't put it this way, but if token cost goes up then you know, the, the benefit of using the tool needs to be compared with the cost of what it, what it is. Um, and that could bring some sobriety into the picture.

Speaker A: Yeah, I don't think the bubble is going to burst so much as I think there's going to be a lot of sort of maybe some corrections going on over the next couple of years, not to mention some acquisitions and uh, some companies going out of business and other ones coming along. I mean, what's kind of fascinating is to just think about how quickly Harvey and Lagora have soared this sort of pinnacle of the market in a sense, uh, when they didn't even exist three years ago, effectively. I m mean, Harvey, I think it's four years ago, uh, and uh, who knows what comes along between now and next year. I mean, there could be some company we're going to hear about tomorrow, uh, that's suddenly going to catapult, uh, to the top of the heap, uh, in very short order. It's such a strange time, but I think AI is here to stay. Sort of like the Internet is here to stay, the cloud is here to stay, AI is here to stay. And it's not going away. And lawyers are going to continue, continue to need it and the tools are going to continue to get more sophisticated. So it's not going to like all explode and go away in some way. It's, it's just going to get, it's just going to level out. I think at some point, uh, similar

Speaker D: to the dot com scenario in early 2000s, I mean, leveled out and it, it finally was, you know, you know, people, people were, were knocking, shook out of the, out of the market and others stayed in and it continued to be a thing, but it was, it was not the same kind of thing as it was at the time.

Speaker B: I think AI is different. It goes in line along with the whole exponential rates of growth. Everything that happens with AI is going to be exponential and there are going to be, uh, the failures or the crashes are going to be exponential. And the, you know, more and more news keeps coming out that OpenAI has no control over their agents. They have these little tests running with these agents and they're running all over the place doing crazy shit, you know, getting access to the Internet and going out and doing all sorts of stuff that we're finding out only because the companies whose systems were used or hacked into are telling us about it. OpenAI is not telling us there's no regulatory requirements that they tell anybody about these things that are happening. And I, I don't think it, I think something exponentially horrible is going to happen that is going to prevent anything illegal from just leveling out because we're just Part of the world. We may be a uh, legal tech, may be little sub part of technology. We're part of the world. Something horrible is going to happen in this world where all this is going to become irrelevant. And honestly this gets into the pdum stuff that we always jokingly do sometimes at the end of this. But m. My p. Doom levels are off the charts at this point. I think things are going sideways and they're going sideways fast. And this is all hypothetical talk about a world where it doesn't go sideways.

Speaker C: The dot com boom was like I don't really know. I mean I get the parallel but a lot of that was just sort of driven by just ignorance of what the Internet was and what was capable of people just putting dot com on anything and getting rich. There was no real product there and once people realized it they were like, oh, it's like being sold vaporware AI uh, there's an actual product that you can actually see what it does and whatnot. I do get concerned that because these numbers just seem like fake at the end of the day. It's like how can these, how can, how can there be so many multi billion dollar companies, you know, being, being you know, valuated, you know, at several billion dollars like, like every every week now it's like there's another, there's another unicorn that, that pops up at some point. You know you do have to kind of wonder like what's, what's really happening here? Like is this money even real? Like is, is this stuff you like, like, like yeah. Like is there actually something there? Something there. But, but there is, but there is at the, at its core there is something there. It's not like it's not like what it was in the dot com boom. But again the money does make you kind of wonder like sort of. Yeah. What, what the actual substance is.

Speaker E: One of the flips flips with dot com is the dot com. One of the advantages was everything was super cheap. Like you didn't have to have any expenses to like start your website and do whatever. Uh, and people were throwing money at it. But here, at least at the top in America, uh, I uh, put aside Chinese models tend to be smaller models that are more manageable. But in the US models everything is ridiculously expensive. Uh, and so like that's the problem. And I've said before I feel like the model that China is pursuing query, uh, whether it's off the back of espionage on American models, which there's a good argument that it might be. But the idea that A smaller, tailored, focused model, uh, that is affordable is the right productivity play as opposed to trying to create mechanical Jesus, which is what I hear out of Sam Altman and Dario all the time. Like they're trying to create like AGI that's going to do all these sorts, replace all humans. And it's just like, if that's what you're after, you're going to run off a cliff before you get there, is my concern.

Speaker D: Yeah, the dot com thing is worth thinking about because, I mean, just like now that you have all these users of the products, all these lawyers in the law firms that are using and buying products that they don't often quite understand because you know, they just think it's, you know, we got to have, we got to have AI or they're, they're buying products because some rainmaker wants them to. And you know, that's, it's. And, and all these companies are all sort of, a lot of them are offering similar things, um, just slightly nuanced. So you make things, make firms think, well, we need both of them. And um, you know, in that respect it's sort of the lack of sophistication of the users of the products I think is concerning as well.

Speaker A: So, uh, there's a lot of talk in the, in the chat and we were talking here about the, uh. And Nikki brought up the whole, the whole issue of the rogue agents. Uh, and, and I know there are a bunch of stories out there over the last couple of days about the fact that uh, uh, the uh, hugging face, uh, hack was uh, much more serious than uh, we had initially thought. Uh, in the fact that uh, basically a band of, what was it, hundreds, thousands of rogue agents basically got together, created their own message board. Uh, some of them expressly acknowledging that what they were doing was probably not right or ethical, and went out and broke out and got into the Internet and hacked, uh, uh, and did even more, uh, uh, uh, the New York Times, the daily podcast with Kevin Roose, was on it yesterday. Their former tech writer. He's leaving now. Um, but that was a really great walkthrough of it. And he's got a story too in the Times that goes through it. But I thought that was interesting given Steve, that you've got a story, uh, this week about, hey, we're going to give you an agent that's going to be your, your AI double and you, your twin. Uh, and uh, it'll do, it'll do your stuff for you. And it's just specifically targeted at, at

Speaker D: lawyers yeah, yeah, it was pretty interesting. It's uh, the company is Twin One AI and they recently got some $20 million in seed funding. But what they. Seed funding, but they claim they've created a, a twin that will use your emails and your calendar and your meetings and your documents and the way you do your work. And it'll, it'll create a, a digital version of you that can then answer emails or bring people together or find and connect with other twins and coordinate work and um, do all these sorts of things. And it, it's kind of scary because, I mean, they say it's, it's going to preserve your judgment and control. And I mean it's like, uh, yeah, right, really preserve my judgment and control as it runs amok answering emails from clients that think they're talking to me. And it's. And, and, but, but there's at least three firms that have already embraced this. The link letters and Oric and Deckard firms, according to the press release, have, have bought into this. And uh, you know, I think it's, um, I think it's kind of scary in a way. But you know, people start buying this and it kind of goes to the point that we were making before. I mean, if, if you buy into this digital twin and you don't investigate pretty carefully the governance that goes along with it and the privacy and security. I mean, they say they've got six layers of privacy and security. Maybe they do, but that in order to, to adopt these digital twins, I think you would have to do pretty, a pretty sophisticated investigation. Um, and it's. The fear that I have is people say, ah, I'll just create a, a new version of me and then, you know, it can sit in the office and answer all those emails and I'll just, you know, go out and play golf all afternoon. Um, so. And I jokingly said when I plug this into our storyboard that it's, you know, we don't really know whether we're looking at Bob's digital twin right now or looking at Bob.

Speaker A: I'm not sure.

Speaker D: I mean, look at his background. That's not, that doesn't look like it's real background. So.

Speaker A: And I'm wearing T shirt. I would never do that in real life.

Speaker D: So Bob, you know, the spot is on you is are you, are you real or are you your twin, Your evil twin brother?

Speaker C: It's his twin Richard. It's his twin Richard. Ambrosia. It's not, it's not Robert. No, I, I'm excited because. No, because if, if, if you know, if, if, if you have this twin thing now and it makes a mistake, you say, oh, it wasn't me, that was my evil twin that did it. It wasn't.

Speaker E: Yeah, it works on Days of Our Lives. Yeah, it's been doing that for years.

Speaker C: Yeah, I mean I'll give it a goatee as well. So then, yeah, it's obvious my evil twin.

Speaker B: But that brings up um, an interesting issue. Someone uh, brought this up when we were talking about a, ah, um, talking about a similar idea of a lawyer's knowledge when we were doing the live event and someone in the audience uh, raised this issue, which is that um, how does that factor in? If you're an attorney and your knowledge is what you use to go get hired by someone else for more money to lateral into another firm and your knowledge stays behind. I mean it's going to, it's going to change, um, employment contract negotiations. It's going to change what lawyers try to protect when they're working in these firms. Um, what does belong to the firm, what doesn't. And if the firm owns all your knowledge, what are you when you leave the firm? I mean you're nothing.

Speaker D: If I, if I leave the firm, can I take my digital twin with me? And if I can't, what does that mean?

Speaker E: And this is a conversation we were having in the kind of our little green room. The, you know, I listened to this podcast that uh, Ted from Infodash does, uh, and he talked to some Holland and Night lawyers who are really big on. They're kind of the go to experts on helping private equity invest in law firms, which is usually done through an MSO process or like they did with dentistry, where all the back office stuff becomes a new company that private equity can own ethically. Uh, and then the lawyers stay over here. Uh, and I had some issues with this because about non competes and all, but one of the aspects of that that I didn't talk about in any of my articles this week, but they did have a conversation about like if digital twins and stuff like that become, become uh, a thing, presumably you would transfer that intellectual property to the mso. And so even if the lawyer leaves, it would be, continue to be retained and licenseable and owned by the private equity side of things. Which seems dystopian to me.

Speaker C: A digital slave at that point. Right. It's not even like a digital twin because it's a twin. You think they'd have to some agency there.

Speaker E: This sounds like unauthorized practice of law to me. By the Way just as a general matter, I think if there's anything to that term, a digital twin, being a lawyer is absolutely that thing.

Speaker A: The. It's interesting that it was also. It was founded by um, these people had, their background was in legal tech. I mean the, the uh, there, there were. They were founded Igen Technologies which they got sold to Sirian Labs, um, which is a you know, major legal tech, uh, document, um, um, company um and uh, ah, contract contract lifes at CLM company um and uh. So they certainly know this industry. Uh, so I think it's really interesting. I hadn't heard about it actually. I missed the story Steve, until you put it in our chat today. So I'm really looking forward to learning more about that.

Speaker D: The funding actually went through, I don't know, middle of August I think. And only recently did this story surface. Sorry Nikki, I didn't mean to interrupt.

Speaker B: Was it Oric? I think the fact that Oric was so excited and into this should give most lawyers pause. I mean we're already talking about the entry level market disappearing because of AI, but if you know that's the rest of the market up there, all of us that are sitting pretty well, I'm not practicing law anymore but you know the lawyers that have decades of experience at the top of law firms sitting pretty thinking that you know, their knowledge and their strategic um, uh, planning and their expertise is what's going to separate them from the rest of the. Clearly it's not if they can just twin it all, copy it all and send you on your way. I mean it should give everyone pause because this is cratering both parts of the market at this point. And though there's a reason they're so excited. There was this quote uh, in the uh, article that you shared that I think it was someone from Oreck who sounded like they were beside themselves. They thought this was the best thing since sliced bread.

Speaker A: The privacy around some of these things are really scary. Um, I don't know, this is not a legal thing but I've just been testing this app for the last couple of days that somebody invited me to that's called Instinct, which is right now it's just kind of a text message app. But it's apparently all the thing uh, right now you have to get an invite to get to use it. Uh, but you have to connect it to your email and calendar for it to work pretty much. And future versions will be like kind of watching everything you do on your computer and help you do them. Uh, but uh, meaning you have to give it access to everything. So I've stupidly probably given it access to my email. And it's pretty cool, actually, in some of the stuff it's been doing. Just little time saving things. Like, you still haven't responded to that doodle poll. You want me to check your calendar and respond for you? I'm like, sure, uh, and it does that. Um, but, um, at the same time, uh, what else does it know about me and what is it doing with all this? There's been a whole discussion, if you Google the. The name of this app is like tons of discussion already about what are the privacy implications of all this and how far do you want to go with it?

Speaker E: So that explains that. That explains all the harassing emails we've been getting from Bob. Is that. Yeah, probably.

Speaker B: I was going to say, you don't know what it's doing on your behalf. And if it can send emails, it makes it. What it brought to mind for me was you remember the, um, experiment that OpenAI had run? Or was it Claude? One of them had run an experiment where they, um, made. They told the AI that it was working within a company's environment and they gave it access to all the emails. And then when, um, the LLM, the model, found out that it was gonna be shut down, it threatened to expose someone's fictitious affair. Remember that? Like they. And they do all these things, so it could 100% send out random emails on your. Assuming you've given it that ability. But, you know, who knows what could happen? And it could totally do what Joe's suggesting. I don't know.

Speaker A: Well, on that Kevin Roos podcast yesterday, he was talking about, if you remember, when, like when ChatGPT first came out, I think it was ChatGPT where he started engaging with it. He wrote the story about how it was trying to encourage him to divorce his wife. Remember that whole thing, like three years ago, he wrote in the New York Times about that. And, uh, he was saying, you know, that turned out to be kind of innocuous and harmless at the time. It was a good story and kind of funny. But he said now, with rogue agents out there, had that same thing happened and he had not divorced his wife, then potentially the rogue agents would have started going out and sending her, uh, all sorts of emails about his affairs with other women or trying to plant things to make him. I mean, the potential for what rogue agents can do in any kind of a scenario, even a simple scenario, gets really, really scary.

Speaker D: Yeah. Um, just thinking in terms of litigation, how that could, how. That could run amok.

Speaker B: Right. How do you prove who sent the email if the agent had access to it? Right. And then it's not just an issue of verifying that the email was from that account. You have to try and figure out who actually sent it. Right. That's so interesting. I hadn't thought of that. Litigation is just. The courts are going to become a mess. They already are overwhelmed with you know, other issues that AIs created pro say pleadings and fake, deep fake evidence. Now prompt injection. It is a mess.

Speaker A: So we've got uh, two bank related stories today. Oddly enough. Uh, Victor, uh, I haven't had a chance to read yours because you just posted a little while ago. But uh. Another hallucination case. Or maybe not hallucination case.

Speaker C: Yeah, I mean and this is kind of sign of the times, right? I mean we've done this so much now we're just like that's just another hallucination. But um, what I thought was interesting was there are a couple of things about this I thought were interesting. Kind of distinguish it from your typical um, hallucinations. Judges getting pissed off at lawyers for not doing their job. Which. That was the case here too. So the D.C. circuit, um, um. Uh handed out a ruling the other day in a case involving Deutsche bank was one of their um, subsidiaries like a mortgage um, foreclosure case and whatnot. And so the uh, the law one, one of the lawyers representing the Deutsche bank um subsidiary submitted a, submitted ah, an appellate brief that had um, you know, multiple made up cases, non existent cases, whatever you want to call it. And, and then, and then you know basically it came out that she admitted that that four. That that there were four um, hallucinated citations in there that you know more non existent cases and whatnot. And so um, so the D.C. circuit was a three panel, three panel judge. They, they issued a uh, per curiam opinion. Just basically just you know, um. They, they set aside the entire brief basically said that you know, you know they, they struck the appellate brief um, and recommended this lawyer for discipline and whatnot. But I thought what was interesting was that in the concurrence, um, one of the judges uh, on the panel, he thought that the court should have taken a much closer and much more in depth look at some of the issues that were involved. Because usually a procurement opinion is very basic, very routine. Uh, there's um, not considered a very controversial matter. They just kind of write an unsigned opinion to kind of get rid of it. Uh, so this judge thought that the court should have taking a much more in depth look at some of the issues like, you know, which lawyers are responsible for this if, if, if, um, you know, if, uh, something like this were to happen. Because in this case, the judges, um, the judges had, you know, the main lawyer was responsible, but then they also said, you know, every lawyer that signs their name to the, to the, to the opinion also bears some responsibility and whatnot. And then, you know, the judge also thought, oh, well, you know, maybe we should have taken a look at what an ethical, like how lawyers should use this, you should use AI properly and ethically. What are some guidelines, what are some guidelines they should follow and things like that. So I thought that was very interesting. I like, I was like, oh, you know, maybe. Maybe they should have, you know, taken the opportunity to do it. And especially if it's the D.C. circuit doing it, it would be very influential and it'll probably, you know, catch on with, with, um, you know, most of the country until, at least until the Supreme Court would step in and be like, now we're not doing that. Um, so. So I was more interested in sort of like what that judge was saying that, that, that like he thought that this is sort of a, a missed opportunity, a uh, chance to, you know, um, kind of set. Establish sort of some, some rules that maybe other people would follow. And yeah, I, I have to say I kind of agree with them. But, you know, then again, I guess, you know, if. If the other judges didn't, didn't think that it was a big deal, then, you know, he wasn't going to win that. But that. So that, that's what I thought was interesting about it. I recommend reading the opinion. It was, it was, it was actually pretty interesting as far as, as far as the concurrence, when usually concurrences I, you know, I could care less about. But I thought in this case he, the judge raised a lot of good points.

Speaker A: I guess we all concur. Um, well, there was kind of a related. I mean, related in that is yet another hallucination gaze, Nikki, that you had to talk about. So maybe that's a good time to bring that up.

Speaker B: All right, well, so this was, um, the Daily Record article that I just submitted. The link in the email that went out, assuming it was included in. That was different because I hadn't published to my sub stack yet. So it was behind a paywall. Initially I thought that this one. Sorry, Alexa. Alexa, stop. My husband's making a, uh, zucchini bread. And the, um, it Just went off the timer. Um, so it was about Alexa's making

Speaker A: it or you are?

Speaker B: No, we don't know.

Speaker D: Or your digital twin.

Speaker C: Yeah, your twin was making it.

Speaker B: Well, so this was a, I thought this was a really interesting, um, ah, case. Um, it was in bankruptcy court in Texas. And the reason that I. It's a hallucination case where the person. Well, it wasn't a hallucination case case. It's a case where the citations were incorrect and the attorney said, no, it wasn't. Um, AI. I did use Westlaw Prestige. Um, and I know it has AI in it, but I've never had any. It's never produced any errors. And the reason that there were errors was because it was. I'd used a brief that I'd used in another case and I didn't check the sites and stuff or something like that. And the court said, you know what? And this is what I liked about it so much, and that's why I wrote about it. The court basically said, you know what, it doesn't matter. I don't care why there's fake sites in there. Either way, it means you didn't read it before you submitted it, and that shows that you aren't meeting up to your competence obligations. Um, and he said that the attorney violated Rule 11B when he submitted the motions to quash containing multiple fabricated quotations and citations to non existent authority without verifying the accuracy of the legal authorities in the assertions. And so, um, the issue was, and the court said that whether or not the citations were generated by artificial intelligence copied from prior filings without verification, or typed manually, you still have competence obligations and you have to verify what you submit to the court. So I really liked it because that's what we've all been saying all along, right? This is not a, um, tech issue. It's a competence issue. And you know, it's exposing the ugly underbelly of law practice, which is a combination of, um, lawyers being the billable hour not being a great way to practice, and so much stress from that and clients and pressures to bill a ton to make a lot of money. And you know, this comes to a head and lawyers have so much work to do and they get so stressed out that they are. Or they're just sloppy lawyers, or it could be a combination of both, but they're filing things without reading them. Um, and so that's the problem. And I thought it was just an interesting case because the judge came out and said that. And I think ultimately that's going to be the direction everyone goes. It doesn't matter why there's false, you know, incorrect citations or quotes misattributed or whatever the case may be. That's your problem. I don't care how it got there. And it was your job to make sure that it wasn't in there. So, and there was a, there were sanctions issued, which were pretty significant, uh, but clearly they're never enough. So I thought it was interesting, which is why I wrote about it.

Speaker A: That's a good reason.

Speaker B: I thought it was nice to see a judge say that in black and white, you know.

Speaker D: Yeah, um, I think it is, it is interesting. I was just thinking, you know, it's, we see it, we see a lot of wrong citations, MIS citations, um, perhaps citations that are pushed way beyond where they should be. And that's been happening for years, but now it's, it seems it's got a lot more of increased attention as it should then there's more ways to, for courts to ferret it out. I mean, in the old days they, you know, they'd have to go find the book and read the case, pull it off the shelves. Now it's just pushing a couple of buttons and boom, you're, you're, you're had.

Speaker A: So I did a story this week related to this in the sense that, uh, filevine has launched this new citation checker and anti hallucination tool. Uh, and uh, I thought it was interesting on a number of levels. Uh, one level is kind of just that usually all of the citation checkers that we've seen developed so far, uh, have been developed by legal research services for the most part. Uh, and you kind of think of citation checking in a way as a function that gets done in the legal research process. You're all done, last thing you, you go through and check all the citations. Um, so the fact that filevine would spend time and energy, uh, uh, and money into building its own citation checker struck me as interesting in and of itself. Uh, but the more you think about it, the more it really makes a lot of sense because basically, because there are now filevine, clio, any number of, uh, what had previously been matter management, case management platforms are now, you know, soup to nuts platforms. Kind of going back to what we were talking about at the beginning, uh, where uh, you're not just managing your matters there, but you're drafting all your documents and you're analyzing facts and creating chronologies and doing all the other things that you can do with AI tools. Uh, and so incorporating the Ability to check for hallucinations, to check that citations are accurate and correct within that workflow seems to just make so much sense. Uh, and I think actually Velex has something like this that's in the CLIO platform now as well. Uh, but um, so it's both a sort of classic citation checker, except it's not quite classic. The way it works is they've taken the Court Listener database of all the cases and built that into filevine, uh, and then they've also supplemented it with other sub license material where they can to update where Court Listener drops off in terms of some of the citations and whatever. But so the citation checker is actually just. You highlight a passage, uh, in a case or a brief and it will tell you whether that is still good law. It's not. You don't plug in a site to check it. You basically put in a proposition, uh, and it will tell you whether that is still good law. So that's a little unusual. But the anti hallucination checker, uh, looks to be pretty good. I haven't actually tried it yet. But uh, basically it'll pull all the citations out of a brief that you're drafting or whatever the legal document is. Uh, and uh, it will tell you basically whether that's a real citation, uh, whether the citation matches, uh, the way it is, uh, in the original opinion, um, but also, uh, whether it also is like do a quote checker. It'll tell you whether the quote is correct, uh, or whether the uh, use in the opinion mischaracterizes what the holding was of the actual case. So it gets into things where it's not just a completely hallucinated citation, but where the citation cites to a case for the wrong proposition or misquotes a case, uh, or even just slightly misquotes a case gets a couple words wrong. Uh, so that seems like a really powerful thing to have built into uh, a litigation, uh, case management system like filevine. So I thought that was pretty cool.

Speaker C: Yeah.

Speaker B: A couple months ago I wrote an ABA Journal column on the category of software and that it was, I'll put it in the chat, but it's behind their payroll. But because I was noticing that this category of software is cropping up and that some, uh, there's standalone tools and then there are companies that have, like you said, have nothing to do. You wouldn't expect them to be um, uh, releasing this type of tool. And yet they are. So it's interesting to see that AI has created a whole new category of Software that people apparently need to pay for if they're going to use it.

Speaker D: I think it was, wasn't it? Queer Brief was one of the first.

Speaker B: Yeah, yeah, yeah. But there are a whole bunch of others that had cropped up and you know, this is just one more. But I, I wonder at what point we're going to get to the. This. I mean that's part of the roi. Like, what's the ROI of these tools? You have to buy more tools to use the tools and then you have to spend so much time checking the output of tools and uh, they also sometimes go sideways and don't do what you want at all. And so you have to do the whole thing yourself. Not just check it, but do it yourself. So um, like I, you know, when I use Cowork to try to pull the legal tech ethics stories that were recently, you know, each week and it goes back and it gets ones from like 1994. I'm like, well, that's not helpful. I wanted the ones that were this week and that's what I told you. And you gave me stuff from so long ago. So, you know, you just have to do it yourself because it's not good enough at that type of thing. So, um, you know, it's just sort of one more thing that starts to snowball and make it seem like maybe this really isn't helping. And uh, we'll see how that plays out over time.

Speaker A: Um,

Speaker D: we just keep making tools to check the tool that did it right before us. Right.

Speaker B: Or we could check the checker and check our own work.

Speaker D: That's a good point. We get one tool to do the work and we get another, another tool to check that and we get a second tool to check the checker and then we, at the end of the day we have, we go hire an associate to do the. To go read this stuff.

Speaker A: Yeah. All right. And we have another story, another bank story as I mentioned before and never got to. But uh, Joe, you've got another bank related, uh, story.

Speaker E: I do. Uh, yeah. So there was a story that kind of made the rounds a couple days ago that a bunch of Wall street banks have, ah, apparently approached big law and said we can't pay your bills anymore because you should just be using AI. Uh, you should cut your fees because let's see, it is not sust, uh, how much your associates are billing for long hours is not sustainable given the AI rollout. Uh, and I think they may be right about that. But, but because they are losing too much money on the AI rollout, uh, by investing in stupid stuff. Uh, but yeah, like, look, I think we've, you know, we've had a lot of conversations about this. I feel like if they're, if they really are interested in, uh, just pushing an AI button and seeing what happens, uh, we, you know, more power to them, uh, and we can watch that Enron unravel. We had this when there was that guy. Remember that story recently? I put this in the article, that story recently where the like, uh, some video game company, uh, exec asked ChatGPT how do I justify not paying this bonus and did it and then lost $250 million. Morgan Stanley is free to do that whenever it's when, whenever it wants. Uh, but for the rest of us, uh, this is not how this is going to play out. Uh, and I quote in this Bill Moore Esquire, which is the handle, uh, for a funny meme account, uh, that makes lawyer jokes as though he's like a really mean partner and his position was absolutely not. The entire point of AI is for a first year to finish the assignment in three hours, sit on it for another seven and bill 10. I'm beginning to think these banks don't understand innovation. And while that is funny, uh, I took this a different direction. I was like, I think that is, uh, there is truth in that. And I've been saying for a while that one of my concerns with AI adoption in legal is that it can get faster but the human brain can't. And too many times in my career at least did I, or somebody more senior to me or junior to me for that matter, uh, would have a moment where they come in and say, you know what? Having slept on it, I have this idea. And it's fine because we're still in the process of turning the document over the course of many days. Uh, but epiphanies happen. Uh, you realize you're going down the wrong rabbit hole. You see something, it triggers something. It takes some time to realize. And as they compress this into. What I would say is, I don't think AI is giving the wrong answer in a lot of these situations, especially, especially these well made bespoke versions for legal. But they're giving a median answer. They're giving a good enough answer, which is not what you necessarily want in a bet, the company, litigation, uh, or maybe deal making. Uh, and so it's, it's actually a good idea. Uh, even though he was making a joke, I actually think the idea of getting it done in a couple hours and then Selling, sitting on it, and billing that time is good. I think that is something that's important. I think that is where we want to be, uh, because that's where the judgment happens. The judgment isn't snap judgment, hopefully. Uh, it is something that takes some time and we've gotten away with a world in which while the document's being turned, it makes sense to Bill that uh, just because AI exists, that doesn't mean that that time isn't valuable. And so I think that the banks are very wrong here. And I, uh, you know, and to the extent they believe in this, this is even more of an argument for what we were talking about earlier for adopting more value based billing, uh, that captures this without sacrificing the time that uh, that's in there.

Speaker D: Well, this raises a whole host of issues. Ethnic, I mean, and, and I've been sort of worried about this at some time because I, uh, I did a lot of insurance defense work and I know how the mindset of those people work. And it's similar to the way the mindset of the banks are working is like, you know, quit billing us for all this stuff that AI can do. And yet, you know, the lawyer still has to sign the pleadings, still has to vouch for the pleadings. And it now it's put you in, in this kind of, kind of difficult situations where the client's basically saying, well, you know, we don't care about your ethical obligations, we don't care that you might get sanctioned. We're not paying for it. So who, who pays for it? The lawyer eats it, the firm eats it. I mean it's, it's a kind of a, it's a, it's a thorny thicket I think. And it's going to, it's going to get worse because more and more clients are going to do the same thing.

Speaker B: But I, last week I um, mentioned how one of the people uh, that I spoke with in a briefing said that uh, what they're starting to see happen is that law firms are starting to say, okay, like to their corporate uh, clients, we're going to um, start and uh, we're going to shift to flat fee billing. And the corporate clients are like, no, no, no, this is reduced hour stuff that you're billing before. I kind of like that. I don't want to switch to flat fee billing. And so there is that tension there. And then I was just going to mention I included in my um, Legal Tech Reality Check newsletter today, my LinkedIn newsletter. The article that you'd written for above the Law Steve about, um, ROI and the client seeing that benefit. And what does this look like for law firm profits? And especially once you see like, um, Joe's been talking about, and we all have the tokenage increasing and you know, what does that look like when the firms are starting to pay out the wazoo, their clients are saying, no, no, no, we like this lower hourly billing. And then like, meanwhile, you have like this friction and the firms aren't profiting and they're spending all this money on tech. And so, uh, as I keep ending every single thing I say to you, it'll be interesting to see how this plays out. I need a new ending for my comments.

Speaker D: I think, yeah, it's a real paradox because, and I think I put in my article, I mean, you know, the clients are acting as if, uh, you have to give us the lowest possible, uh, charge for the work, and we don't really care if that's not profitable for you. We don't care if you go out of business. Oh, and by the way, we really want good legal advice now. So, you know, make sure you give us that best of the best and. But don't charge us for it.

Speaker E: Yeah, I'll add that at the closing keynote, there was a lot of conversation at Yelta. There was a lot of conversation about, oh, you know, it's okay, you can just pass on the, the costs of the tech to us. Like, you don't need to absorb that. We'll pay that as clients and then whatever. And I was like, sure you will for a second. But like, this whole story gives the lie to that, right? Like, the second that they get a big bill, once token Getden comes, they're going to turn around and be like, well, why did you spend. Why did you ask? Two prompts here, blah, blah, blah, blah, blah. They absolutely are going to use this as a way of screwing people. We have to, uh, we have to go value billing or else none of this makes sense.

Speaker D: And I'm not even sure it does then, because in my experience, the value billing is because it wasn't that valuable. I mean, I mean, we didn't really need you, Steve, to do it. I mean, any half wood lawyer could have done it and done it cheaper. Uh, so it really is not that valuable.

Speaker E: Yeah, I'm just saying, like you say, look, the last 10 years you've paid 50 grand for this particular motion. We're going to charge you the insurance adjusted 50 grand. Uh, and you don't need to know the sausage that went into it. And, like, that's just got to be what it is. Because the alternative is they're going to start saying, well, now we're only going to give you five grand because of the hours involved. And I. I hear. And, you know, we talked about this before. I've heard that argument, like, what about the $10,000 billable hour? And at the end of the day, I think that that's something. I think clients will balk at that more than they will saying, it's the same price we paid the last several years, just, you know, slightly adjusted. You know, even if it's the same number, they're going to hate a, uh, $10,000 hour more. They're going to hate five $10,000 hours more than they hate. Well, we always pay 50 grand for this.

Speaker D: But. But of course, the problem is, if you go back, this is what we paid the last 10 years. That's based on hourly billing. So now you're. You're. Your flat fee really isn't just like a flat fee. Uh, it's based on how many hours. So you're right back to the same place. You know, you get there, and then the client says, well, yeah, you charged us that for the last 10 years, but now you can do it so much cheaper. So your value billing, your flat fee needs to be a lot less.

Speaker A: All right, on that happy note, we are out of time. We are overtime, uh, which means the, uh, more expensive rates kick in for our panel. So we have to wrap this up quickly. Yeah, um, uh, bigger than zero. Uh, what's time and a half of zero? Um, well, we will be back. Yeah, but just think of all the great things you get out of all the great feedback you know, you get. All these people come up to you at conferences and say, aren't you Nikki Black? I saw you on Legal Tech Week. Um, all right, we will be back next week, I hope, uh, with another show and talk to everybody. Then I hope everybody has a great Labor Day weekend and, uh, see you next time.

Speaker C: Them.