What’s on tap for 2026 when it comes to generative AI?

Episode Notes

Transcript

Announcer:

Welcome to the ABA Journal, Legal Rebels Podcast, where we talk to men and women who are remaking the legal profession, changing the way the law is practiced and setting standards that will guide us into the future.

Victor Li:

Welcome to 2026. We did a predictions podcast last January and enjoyed it so much, we figured we’d do it again. Guess we should have seen that coming. As for things we did see coming on this podcast last year, we predicted that generative AI would continue to make inroads into the legal industry, and that ended up being right on the money. Multiple studies and surveys have shown that lawyers are using it more in their practices and day-to-day operations. We also predicted that the federal regulatory landscape would remain confusing, and that was certainly the case. We also foresaw that more law schools would start to embrace AI, and that was also true. So as such, we figured we’d bring back our chief prognosticator from last year. Oliver Roberts is here once again to talk about what we might expect in the coming year, and given his track record last year, maybe he’ll also give us some lottery numbers to play or tell us who to bet on for the Super Bowl this year.

My name is Victor Li, and I’m assistant managing editor of the ABA Journal, and I am host of the Legal Rebels Podcast here on Legal Talk Network. And as I mentioned, our guest today is Oliver Roberts. Oliver is the co-director of the AI Collaborative and an adjunct professor at Washington University and St. Louis School of Law. He is also editor-in-chief of AI and the Law at the National Law Review and the co-chair of the AI Practice Group at Holtzman Vogel, and the CEO/co-founder of Wicker.ai, a legal AI startup in education and training. He’s joining me today to talk about what we might expect in the upcoming year when it comes to AI and the law. Welcome to the show, Oliver. Thanks again for joining us.

Oliver Roberts:

Thanks so much, Victor. Happy to be here.

Victor Li:

So I gave a very brief elevator version of your bio. You obviously wear a lot of hats. Can you tell me a little bit more about yourself and how you got to where you are today?

Oliver Roberts:

Yes. I like to stay busy and all of it really does just revolve around AI and law. My background is I graduated from Harvard Law School a couple years ago, five years ago now. I’ve worked at two bigger firms. Now I’m at Holtzman Vogel, where they’ve really given me a great opportunity to expand my own professional development and network in the AI space. So I’ve started a startup on the side in Wicker. I’ve had the opportunity to teach at Case Western. I’ve taught at Southwestern Law School, Wash U Law School, many others in the AI space, and it’s such a exciting area and constantly changing. I mean, I teach at one school. In January, I teach at another in February. I have to change 50% of the content. That’s how quickly changing this area is. So it’s very exciting to have a foot in academia on the media side at the National Law Review, as well as in practice at a Holtzman Vogel.

Victor Li:

So I don’t think I asked you this question last time, but so what did you do before generative AI became commonly available to people? What was your practice like and how much has it changed since the advent of generative AI?

Oliver Roberts:

Yeah. So generative AI, it’s been around for many years now, but the big boom onto the scene was ChatGPT. That was back in November 2022. So before that, I was in law school till 2021. I actually practiced Skadden Arps in New York City right out of law school. And I left after nine months to start my first startup. It was a tech platform for hiring, and I outsourced all the development to another country, literally lived in another country to work on the ground to build this tech product. So that’s where I saw firsthand the development. And that was all pre-generative AI. Post-generative AI, boom, where it became widely available and also usable through APIs has really changed a game in software. So I’ve been building in this space pre-generative AI, now post-generative AI, and just the capabilities are so much far beyond what we had in previous iterations of software.

It’s

Victor Li:

Interesting because I mean, look, not to blow smoke up you, but Skadden Arps as an associate, many people would kill for that job. I mean, a lot of people would sell their souls to be where you were. So what sort of pushed you to decide, okay, you know what? I want to go on my own and do something more tech oriented.

Oliver Roberts:

Yeah, it was not an easy decision. At the time, actually, while I was studying for the bar, I was working on a political campaign and I have no background in politics, no connections or anything like that. So it was very tough to get into. It’s tough to get into politics without the right connections. So I kind of worked my way through, volunteered, did a lot of work. And once I got onto a political campaign, I was in charge of hiring interns. So I brought a bunch of other people who had never worked in politics before, got them in the door. So the whole point was not to build a tech platform. It was actually just to build a platform that helps people get into politics. So it was more the mission of that that drove me to leap Skadnorps and build that as opposed to the desire to just build a tech platform.

So I’m very much driven by the core concept of what I’m doing or building as opposed to whether it be related to tech or not. And that kind of translates over into AI right now. I think AI has immense capabilities for really leveling the playing field for people to get into business, to acquire education. And that’s really what drives me in terms of all the stuff I’m doing in AI, whether it be on the media side, giving people opportunities to write about AI, get their names out there through the national law review, or doing it in the context of teaching at law schools where it show people, “Hey, here are the opportunities for you. ” As a young lawyer entering a big law firm, you could be one, two years out, but you could literally lead the AI revolution at your law firm. If you’re engaged, you know about the tools, you could use the tools, help your firm vet vendors and be on the AI committee.

That’s something that young lawyers could literally do right now and make a name for themselves as opposed to waiting 10, 15 years to make partner and then become knowable in that specific legal space.

Victor Li:

Yeah. Yeah. Because I mean, like you said, the firm model has been in place for how many decades now. It seems like this is a good way to cut some of those years out or get close to the front of the line. It’s at least a pretty good way to disrupt it, right?

Oliver Roberts:

Oh, certainly. I mean, the reality is I started my journey in AI about a year ago, third year associate at the time. I’d basically be completely irrelevant had I not put the effort in to learn about AI, get involved, teach at schools, write about it. So this is very much an area where you could step in. It takes a lot of work, especially day-to-day, just keeping up with it, but there’s really no reason why a young attorney could not get out there and make a difference in this space that they put the time and effort in.

Victor Li:

Gotcha. All right, cool. So let’s dive in. So obviously generative AI continues to dominate the legal tech sphere. Do you think we’ll see more of this in 2026? Do you think we’ll see more adoption? Do you think we’ll see more deals that are centered around AI? Or do you think things have hit a ceiling a little bit, like the adoption rate has kind of plateaued a little bit or maybe some of the hype is dying down? How do you see the year playing out?

Oliver Roberts:

I think the hype’s going to die down. However, that doesn’t mean the usage and adoption of generative AI will die down. It’s really becoming commonplace in a lot of the stuff that we’re doing, a lot of the softwares we use. I think we take it for granted and the capabilities are still improving exponentially. So we’re not seeing a drop off in generative AI performance or model collapse or anything like that where some people prognosticated that we’re going to run out of data and then some of these models are getting trained on synthetic data and there’s going to be this degradation and quality. We’re not seeing that. So we just see an increase in performance. We see competitors entering the market. Google is now neck and neck with OpenAI, whereas Google from the outset in 2023, they were well behind OpenAI in terms of their model performance.

So that increase in competition, increase in quality is driving costs down. To use an API that is as in a developer to actually use an API from OpenAI or Google, the cost has dropped exponentially. So these models are better and it’s cheaper to use. So I do not think we’re going to see a reduction in generative AI use in legal tech products. And I don’t see it going away in the legal field either. These tools are only getting better, not worse.

Victor Li:

Are there any growth areas within the legal field, like either practice areas or practice management, things that they could be doing that lawyers maybe aren’t doing as much, but they could be, and you think maybe they might start doing in the coming year?

Oliver Roberts:

I think the areas that we’ve seen so far is, of course, an e-discovery and contract review. That’s where generative AI has been more widely accepted, more so traditional AI that is just especially new discovery with TAR dating back to 2012. Some untapped areas right now in the legal profession is the judiciary. I think we’re going to see in 2026 a lot more experiments by judges actually using AI. We now have at least three issued opinions by judges in which they’ve actually used chatbots to help interpret different provisions, whether it be in criminal statutes or just interpreting sufficiency standards. That’s one from Ross v. United States and the DC Court of Appeals. So judges have already publicly experimented with these tools and I know of other judiciaries, some of them working with directly, they’re actually testing our products from big legal tech companies to see how they could automatically generate bench memos and summaries of cases just by inserting the public pleadings into those tools.

So I think we’re going to see a lot more experimentation in the judiciary. But with that said, I also think we’re going to actually see a lot more judges mess up in terms of issuing opinions with AI hallucinated cases.

We have two very high profile cases, Judge Neels over in New Jersey, federal judge, as well as Judge Wingate down in the Southern District of Mississippi. His was quite an egregious one and I’m hot take, but I’m pretty surprised impeachment proceedings were not brought. That was a case from the summer. His opinion, the very first page in a footnote, he listed plaintiffs that were not even in the case.

He

Listed defendants that were not in the case. He cited the wrong statute. He misquoted the record. He introduced new facts that were not even the case. It was the most shocking misuse of AI that I’ve seen, whether it be a lawyer or a judge. And I do think we’re going to see more misuse by judges more so vicariously, it’s actually going to be their clerks misusing it and just a lack of oversight by the judges.

Victor Li:

Maybe it’s one of those … Well, what”s the term that’s like not skunkworks, but you know how when you do something deliberately wrong as it’s a way to prove a point, maybe that was it, right? I don’t know.

Oliver Roberts:

Yeah, perhaps. We’ll give him the benefit of the doubt. Maybe we’ll say that that was the reason. Maybe he was just giving us content for this podcast. Sure,

Victor Li:

Sure. Right. Well, yeah. That’s what it was. Yeah. He’s a fan of the show.

Oliver Roberts:

Exactly.

Victor Li:

So what about Agentic AI because that’s the big buzzword right now. Everyone’s talking about it. First of all, I guess, how would you define it?

Oliver Roberts:

So Agentic AI, it was a hot topic even January of last year.

And I do think it of course has its own technical definition, but I do think it’s more of a marketing thing. We moved from AI, boom, then there was that clarification of, oh, it’s actually gen AI that’s changing legal profession. Then it was agentic AI. So I do think it’s more of a marketing flash than anything. From a technical side, what it is is just autonomous AI systems that can actually plan and execute different steps autonomously. So it’s not a lawyer uploading a contract to ChatGPT and saying, “Hey, pull out all the important terms.” And then the lawyer has to type in again, now improve the provisions related to non-competes or NDAs. So basically these are autonomous systems where you have a plugin. Spellbook’s an example of one that I know is agentic capabilities where you have this contract review plugin in your Word document where you could actually execute functions that happen autonomously.

So you could lay out what they call a playbook. You could say, whenever you’re reviewing a contract, ensure that these key terms are in there. If there’s any issues regarding a non-disclosure provision in there, then it’ll automatically find that non-disclosure provision, redline it out, insert the new one. So it’s actually executing many different functions independently and autonomously. So that’s a big difference. LexisNexis Protege has introduced agent capabilities now where you ask a research question, it’ll actually go step by step pulling from primary sources. Secondary, it’ll actually internally check its own accuracy before you see an end result. So there are legitimate technical differences between a gentic AI and traditional generative AI using a chatbot. However, I do think it’s more of a marketing thing and legal because the reality is lawyers just want tools that deliver efficiency. Whether they know it’s AI, gen AI or Agentic AI to the lawyer, really doesn’t matter.

Is it going to deliver better quality work product? Will it deliver more efficiency? Is this something that the client actually wants? That’s really the core question. And a lot of the other stuff is a little bit of marketing buzz.

Victor Li:

Gotcha. But let me ask you this though, because whenever I hear agentic AI defined or described to me and that autonomous quality is emphasized or mentioned, that automatically kind of provokes a reaction from me. And I can imagine that that would provoke a pretty strong reaction from lawyers who are known to be risk averse, known to maybe want to have more control over things and maybe aren’t going to want to relinquish that control because of what could happen and blah, blah, blah. So do you think that that could inhibit maybe some of the adoption or the embrace of Agentech AI or do you think the, like you said, the efficiency and the ease of it is just going to be too much to resist?

Oliver Roberts:

So in theory, yes, just hearing the word autonomous and delegation of basically just a lawyer’s judgment sounds like it brings off alarms and concerns for lawyers. In practice, I don’t think it does because the reality is if I’m just going to go on Westlaw and search a question in the traditional search box, like what are the elements of negligence in New York State? It’s going to give me an answer. If I go use Lexis Protege’s agentic capability, I ask the same question, it’ll run through internal processes autonomously, but the end result, me as a lawyer, I’m still going to look at both end results and use my own judgment if that’s quality or not. So yes, the functionality that runs could be agentic and autonomous. It would not use lawyer decision making, but lawyers are still going to be reviewing that work product that comes out the same way they always have done so, even using traditional search features in Lexus and Westlaw.

So theoretically, yes, it sounds concerning that this stuff’s running autonomous, but a lawyer is still at the end of the day going to review the output from that autonomous process and use their own judgment.

Victor Li:

Gotcha. Okay. That makes sense. Before we continue, let’s take a quick break for a word from our sponsor. And we’re back. So in this segment last year, we talked about regulations and whether we might see something from the federal government, and we did kind of. In December, President Trump issued an executive order claiming to preempt state regulations and stating there needed to be uniform federal standards. So putting aside the constitutionality of the order and whether or not you can really preempt state regulations, do you think there needs to be a federal standard?

Oliver Roberts:

Definitely a loaded question. So I am of the camp that 50 different potential regulatory regimes throughout the country would in fact inhibit AI development and expansion and set us behind other countries, mainly China, who is up there in terms of model development and performance.

As for a federal standard, I mean, Congress could in theory, they have many different options when it comes to federal preemption. You could have a federal preemption happen through setting the floor, basically saying that there could be this federal standard that is the bare minimum and the states could actually regulate above that floor. The federal government could step in and set the ceiling. That is, this is going to be the highest uniform standard for AI regulation and no state can go above that. Or Congress could step in and preempt the field and actually have no regulation step in. They’ll just say, “The states cannot regulate in this domain.” So I do think there needs to be a federal standard, but just saying there’s a federal standard, it could be any three of those options. And it could be totally pro- regulation, let’s get the most stringent federal standard, and you’d probably set that as the floor and let the states regulate above it.

Or you could say, “I’m pro uniform federal standard,” and you could take the position that, yes, we’re just going to preempt and the federal standard is just going to be nothing and states can’t operate in that environment. So it’s a little bit more complex of a discussion when you talk about actual the operations of constitutional preemption, but my opinion is that you cannot let the states, at least at the model layer, and there’s another distinction here, where exactly are you regulating in terms of AI development and deployment? Because states could step in and every state could have their own law related to disclosure of training data. And now every single large language model, no matter where they’re deploying in all 50 states, they might have to release training data in California or New Mexico or New York, and there’s more stringent requirements in Florida. So having to put up with that at the model layer will literally inhibit every single downstream AI application that relies on those large language models.

So I do think there’s big concern, especially when it comes to states regulating model layer development that has these frontier models, whereas it’s less concerning for me if you have a state like Tennessee that passes the Elvis Act in which they’re trying to prevent AI from mimicking people’s voices, mostly musicians. So they’re able to actually profit off that to musicians, as opposed to having to compete with AI that’s taking their name image and likeness. That would be more of a downstream, you’re regulating applications that employ AI for nefarious reasons. So there’s a distinction not only in terms of the constitutional operations of federal preemption, but also in terms of at what juncture are you regulating AI applications?

Victor Li:

Let’s say I made you, or you were chief magistrate of this country or the AI car of this country. What standard would you propagate or what do you think would strike the right balance?

Oliver Roberts:

Right now, I would like to see a federal preemption in the form of preempting any state AI regulation that regulates the model layer. What I mean by that is any law, for example, in California, there’s a law that just went into effect January 1st of this year, 2026, that requires high level disclosure of training data. That’s California AB 2013. I don’t think any of these states should be in the business of regulating the model layer. That is what training data goes in, how these models are actually trained. I think states could retain authority over applications like regulating, for example, in Illinois, whether AI should be able to deliver healthcare advice or service mental health therapists. That I think is more adequately in the domain of states, but I do think we should see a federal standard that preempts any state AI regulations and laws that are regulated the model layer.

Victor Li:

So kind of in that, I mean, California is sort of the example that gets held up a lot. And I guess I think Trump specifically mentioned them in his order. What do you think of the argument that if a lot of states just end up mimicking what California does or just kind of follow their standards just because so many companies have to do business with the state and so many of these companies are based there and whatnot, that obviously that they would then kind of adopt that as their own standard for businesses, then that could be workable. Or do you think that’s not a workable solution?

Oliver Roberts:

I mean, it’s quite possible. I mean, you see started with Colorado AI Act. Of course, California is different because all these large language model companies are actually based there.

So these large language model companies will actually need to conform to those California standards. And there’s the argument that if California is setting that policy, then California basically dictates their national policy. And because large language models, especially at the model frontier layer, that is these frontier models at that layer of the actual development of a large language model. And California is basically just dictate international and international policy when it comes to economics, geopolitics, because California could in theory put these really stringent regulations, slow down AI development, that then hurts our entire economy given how widespread generative AI is embedded in everything we do across all our companies, as well as in what we’re doing in terms of national security. There’s a tire race between us and China and whoever is the best technology could ultimately prevail in the long term in terms of geopolitical standing.

So it’ll allow California to just unilaterally dictate what our AI policy is by virtue of regulating the model layer, I think is very concerning. So it’s not just that argument of, hey, fragmented 50 different state regimes that these AI companies will have to navigate in terms of compliance, but essentially what we’re doing is allowing California to dictate nationational and internationational policy at different levels.

Victor Li:

Gotcha. And there’s a political component to that too, right? And so in that vein, then asking you to put your prognostication hat on, what are the chances that we could get something from the federal government this coming year, or do you think that that ship has sailed?

Oliver Roberts:

This is such a complicated, and I’m going to quickly run through. A lot of people thought the federal preemption debate came out of nowhere. It was in the one big beautiful bill act from June of this past year, ultimately passed in July. Obviously federal preemption was in it, but it was taken out 99 to one vote. So this did not come out of nowhere. Deep Seek, that is the Chinese AI model. It’s both the company and the name of the model. Out of China, this time last year, they released R1 and V3, these very strong large language model, it had performance similar to OpenAI. So there’s this big concern. Oh wow, China has these capabilities we didn’t think they had in AI. Then you fast-forward to the Trump administration issuing its first executive order, revoking the Biden executive order in AI. Then Trump issued an executive order in January of last year saying, “Hey, we’re going to make AI a huge national priority.” OpenAI submitted their comments to the federal government related to that executive order saying, “We need federal preemption.” And they said that in March of 2025, and then it slipped into that one big, beautiful bill act, passed the House, but even one Congresswoman said, “I voted for it, but I had no idea federal preemption was in there.” It was a thousand page bill, they slipped it in

And then everyone saw, wow, it’s there. And that catalyzed this huge movement for federal preemption. Of course, it died out after that one big, beautiful Bill Act passed in July, came up again past couple months, and now we got this executive order. So federal preemption didn’t come out of nowhere. It has been this real issue, not just navigating 50 different regimes, but it has always been tied to us versus China. So that’s been a big concern. So that now brings us to your question, what’s going to happen this year? I think it’s going to be tightly intertwined with what happens in China. If China keeps releasing these advanced AI models, that is going to be a stronger argument that, hey, this is a national security concern, an international geopolitical issue. Stuff that we could see happen is China, Xi Jinping and his New Year’s Eve address, this just a couple days ago, he said that it’s their policy that they’re going to reannex Taiwan.

Taiwan is obviously very important for us. It’s the global hub for semiconductors. It’s where we actually build these chips that Nvidia has designed. So if China takes a step toward Taiwan, immediately that’s going to threaten our AI industry. And that’s going to be all hands on deck legislators who are opposed to federal preemption because of different consumer protection concerns. Those concerns are going to go out the door if national securities escalated to the point of China threatening our core location for semiconductor and chip development. So this is not a simple, “Hey, our legislator’s going to change their tune. This is going to be directly intertwined with geopolitical events that are going on throughout the year.”

Victor Li:

No, thank you for that. That’s actually a much more in- depth answer than what I was hoping for because I mean, I think the easy answer would’ve been, “Hey, who knows with this federal government, right?” We’ll go with

Oliver Roberts:

That. Yeah, we’ll go with that too. Who knows?

Victor Li:

I mean, no

Oliver Roberts:

One knows. No,

Victor Li:

No, no, no, no, but I’m saying I didn’t think of it like that. So no, thank you for that. Because the easy answer is always, “Hey, you never know with this government especially.” And asking the Senate to pass anything or the House to pass anything is herding cats. So no, thank you for that.

Oliver Roberts:

I appreciate it. Yeah, definitely. And to your other point, I mean, this executive order that came, we’re going to see how this AI task force created by this executive order, how they’re going to challenge state laws. I mean, they’re going to challenge, of course, the Colorado AI Act, the California Frontier Model AI Act as well, perhaps the one in Texas as well, which would be an interesting change up to actually go after a AI regime in a conservative state. But also the FTC, the FCC, they’ve been directed to analyze what existing statutes on the book on the books could actually give rise to regulations that preempt state AI laws as well. So a lot of people are like, oh, this executive order can’t preempt state laws, but that’s not really what’s going on. It’s going to play out in terms of litigation and it’s also going to play out in terms of FTC or FCC regulations that are not even issued yet.

And they’re going to be derived from statutes that have been duly passed by Congress. So there is the potential that these regulations could be legitimate and could preempt state AI laws. So I think a lot of people will be surprised that this is going to carry more weight than they may think otherwise.

Victor Li:

So your prediction is more litigation, more lawsuits.

Oliver Roberts:

Yes, of course.

Victor Li:

Thanks. Well, hey, maybe those judges can then, who are experimenting with AI, maybe they can take a test drive on one of these cases then.

Oliver Roberts:

That’s

Victor Li:

True. We’ll see. We’ll see. Gotcha. All right. So let’s take another quick break for a word from our sponsor and we’ll be right back. And we’re back. So now let’s talk about a topic that’s very near and dear to your heart law schools. So how are law schools currently teaching or training students on generative AI?

Oliver Roberts:

It is still widely variable. So you have some schools, Washington University in St. Louis, one of them where I teach Case Western as well, where I teach as well. They’ve rolled out comprehensive programs. Case Western was the first law school in the country to acquire an AI certification course. And I taught that last year in February and teaching it again this year in February. So Case Western is well ahead of the ball.

When it comes to AI education, they take it very seriously. I know in their legal research and writing class, they teach their students about it. All the generative AI capabilities in Westlaw and Lexus. Was U is pretty similar in that boat. Other schools that are doing this also is Mississippi College of Law, Southwestern Law School. Santa Barbara just announced an AI education program. So a lot of these schools are now taking it seriously. But with that said, other schools have taken the perspective that, hey, we look at the curriculum, we already have one AI regulation class that’s 10 to 15 people,

And some schools think that’s sufficient. But it’s my opinion that when you teach AI, there’s many different things to teach. You can teach about AI regulation, but that is entirely distinct from teaching about AI tools and practice, actually teaching lawyers, no matter what area they’re in, transactional litigation, teaching them, “Hey, here are the tools out there that are powered by generative AI. Here’s how to use them responsibly. Here’s how to use them ethically.” That is totally distinct in teaching someone about the Colorado AI Act and federal preemption. Those are two totally different areas. Then you could also teach AI ethics usually in an ethics course. A lot of it’s being taught in legal research and writing, just teaching about, hey, here are the ethical responsibilities like ABA Formal Opinion 512 and stateBar guidance. But point being, some schools are really leading the charge doing this comprehensive education on AI.

Others are just doing it in a siloed fashion on a class by class basis, but I still do think that the legal profession is pretty far behind. I did an informal survey of 85 legal professionals and 84% of them, 84% said law schools are either lagging behind with significant gaps or outright inadequate in legal AI education. And I’ll caveat that by saying this was not an even distribution of the legal profession. This is just 85 people in my network, but I think that’s at least one data point to show you how the profession is viewing the inadequacy of law school education on AI.

Victor Li:

I think it jives with a lot of the anecdotal evidence that just from what I’ve heard even. But how are students specifically being trained? Is it mainly just, hey, how to use maybe co-counselor or something to draft a memo or how to use ChatGPT to search through a contract or something? How is it specifically that they’re being trained or is it kind of a mishmash of things?

Oliver Roberts:

Yeah, it depends school to school. The kind of foundational baseline at a lot of these schools is at least teaching students that, yes, here’s Westlaw, here’s Lexus. Traditionally, here’s the methods of legal research, but then also now that you have these capabilities like co-counsel, their generative AI chatbot to do legal research as well and Lexis Protege showing students these tools how to use them. Also, I think it’s very important, and I do this in all my courses, showing students, you could ask the same question to Westlaw through the traditional search box, the traditional method of legal research, ask the very same question in Westlaw’s co-counsel, and you’re going to get different responses. It’s going to give you a different case law. So it’s showing students that, yes, these are new tools, but you can’t just rely on one or the other. You usually still have to rely on the traditional method to verify that you’re actually getting the most authoritative cases and most relevant information.

And that’s also coupled with you teach a law student how to use the tools, you got to teach them about the ethical considerations. You should not be putting client sensitive information into, say, ChatGPT or Clawed free versions, even paid versions are a concern for a lot of law firms.

So it’s teaching students not only what tools are there, but how to use them, how this compares to traditional methods of thinking and reasoning, and then also covering the ethical portion so students use them ethically and responsibly.

Victor Li:

And do you think we’ll see more schools starting classes or having certification programs this year, or do you think schools are just willing to standpoint and see how things play out?

Oliver Roberts:

Schools are moving. I don’t even think it’s a question anymore at the leadership level at law schools about whether they need to teach AI. They know they have to. The demand is there from these employers that are hiring. And I’ll even say anecdotally to call back what we talked about earlier is Judge Niels, the judge who included fake citations in his order in New Jersey, he blamed it on a law school intern. So you literally had a law school intern walk into his chambers. The law school intern used, I believe it was

ChatGPT,

Didn’t know about the limitations, put a case citation into the opinion, no one in the chambers actually reviewed it and it actually got into official opinion and he blamed the law school. So I could tell you that it’s well known now throughout legal academia that you need to teach about AI. The biggest limitation is who could do it. A lot of the librarians are taking this on, but it’s just such a fast moving area. You need to have connections with legal tech vendors. You need to follow the model developments, you need to follow the legal ethics angle of it. So it’s more of a capacity issue as opposed to a priority issue. It’s a priority for a lot of these schools, but the implementation is often difficult.

Victor Li:

I Gotcha. What about for schools, very prestigious schools like your Harvards and your Yales and whatnot. Do you think that they’ll embrace or start moving more in this direction? Or do you think they’re of the opinion that, all right, well, we’re Harvard, we’re Yale, we’re Stanford, whatever. People want to come to us no matter what and people will hire our graduates no matter what. So it might not be as pressing of an issue for them. Do you think that’s at play or not really?

Oliver Roberts:

No, I think that would just be an application of the responsibility of law schools. You have to teach your students about what’s going on in the legal profession and the most modern updates in education. So I don’t think any prestigious school is going to be off the hook of teaching their students about the most recent developments. And there’s still close competition between a lot of these law schools, even in the T14 to stay because there’s been a lot of movement in the past couple years. So I do think these schools are not going to be able to just lay back and say, “Hey, we have a high ranking. We have the name ID. Let’s not teach our students about this. ” So I think every law school is going to have to be mobilized on this.

Victor Li:

Gotcha. What about from an academic integrity standpoint? Because obviously that’s always one of the big concerns for why people are like, “Oh, we got to ban this or we got to not let students do this or we have to be careful with this technology.” And there’s software that purports to catch things that have been drafted by AI or have a significant amount of stuff that came from AI and the software might not be so accurate or might not be very good or some of it might be better than others. Do you think the software will improve to the point that it’ll be able to catch cheaters and catch people who just copy and paste from one of these chatbots with a high degree of accuracy, or do you think it’s still going to be the way it is for the coming year?

Oliver Roberts:

It’s still going to be the same issue. So one of the leading tools used to catch plagiarism in AI, written content is turn it in, but even their tool is not good enough to catch AI. It could give you kind of a point in a certain direction, so this might be AI generated, but law schools cannot act on that alone to tell a student that they violated academic standards by doing AI plagiarism. I do not see this technology changing anytime soon. The only way I actually see this changing is if you actually have OpenAI as well as Google, Claude come together and actually help build an AI water marketing system. One example is Google. They have SynthiD where they actually put water marking on their text. You cannot see it with a human eye.That’s how subtle it is, but it does actually track text. You have issues if someone takes an output from Gemini that is Google’s chatbot and then moves the words around and breaks up that synthetic tracking, but at least it’s better than no tracking at all.

So until these model companies actually step in and create their own AI plagiarism, checkers and trackers, because they get embedded in the text that comes out of their models until that happens, I don’t see any vast improvements in terms of these AI plagiarism catchers and trackers. Gotcha.

Victor Li:

And finally, what about for you personally or professionally? What are some predictions for yourself that you would like to talk about here on the show?

Oliver Roberts:

Nothing really. Just going to follow AI, stay up to date, see what I could do to contribute to the AI space and community. I got some plans I’m working on, but we’ll see how they play out over coming year.

Victor Li:

Gotcha. All right. And if our listeners want to reach out to you to talk about any of these issues, what’s the best way to do that?

Oliver Roberts:

You can reach out to me at Oliver@Wickerd, W-I-C-K-A-R-D.ai. So [email protected]. Feel free to reach out if you have any questions, comments, or just want to chat AI stuff.

Victor Li:

Great. Thanks again for joining us, Oliver. I appreciate it.

Oliver Roberts:

Thanks so much. Really appreciate being here.

Victor Li:

Yeah. If you enjoyed this podcast and want to hear more, please go to your favorite app and check out some other titles from Legal Talk Network. In the meantime, I’m Victor Li, and I’ll see you next time on the ABA Journal, Legal Rebels podcast.

Announcer:

If you’d like more information about today’s show, please visit legalrebels.com, legalTalknetwork.com. Subscribe via iTunes and RSS. Find both the ABA Journal and Legal Talk Network on Twitter, Facebook and LinkedIn, or download the free apps from ABA Journal and Legal Talk Network in Google Play and iTunes. The views expressed by the participants of this program are their own and do not represent the views of, nor are they endorsed by Legal Talk Network its officers, directors, employees, agents, representatives, shareholders, and subsidiaries. None of the content should be considered legal advice. As always, consult a lawyer.

Continue Reading