Episode Transcript
[00:00:00] Speaker A: Imagine a justice system built on rigorous evidence, not gut instincts or educated guesses about what works and what doesn't.
More people could access the civil justice they deserve.
The criminal justice system could be smaller, more effective, and more humane.
The Access to Justice Lab here at Harvard Law School is producing that needed evidence. And this podcast is about the challenge of transforming law into an evidence based field.
I'm your host, Jim Greiner, and this is Proof Over Precedent.
[00:00:36] Speaker B: I'm Elizabeth Guo and I'm a student in the Access to Justice seminar at Harvard Law School. I'm so excited to be joined here today by Professor Ellen Murphy from Wake Forest Law. Professor Murphy, thank you so much for taking the time. I'm really grateful to have the chance to speak with you and I'm so looking forward to our conversation today.
So without further ado to start off, could you please introduce yourself to our listeners and tell us a bit more about your scholarship, including how you've arrived at AI and Access to Justice.
[00:01:07] Speaker C: Well, Elizabeth, thank you so much for inviting me. I am very impressed with the work of the Access to Justice Lab at Harvard and I'm very excited to be here.
My route to Access to Justice and NAI is a little bit interesting. I started teaching the unauthorized practice of law about 15 years ago to working professionals who are not licensed lawyers. So individuals who are working in a heavily regulated industry who need to understand the law, but really as a risk management tool. And I became very interested in UPL and in 2017, I co authored the first textbook on the unauthorized practice of law. And I, I, when I look back over that, like, I don't know, 15, 16 years, I think about how when I was first teaching upl, we really focused on the unauthorized practice of law by a natural person or maybe an entity, you know, a legal person. And then there was a little bit of a shift that started coming with respect to software, and then the software got a little bit more sophisticated and we solved some cases there. And now we are at generative AI, of course. And so, you know, for a long time this has been about UPL as it relates to persons. And now we're really, we, we've introduced this new thing or quasi person, I guess some people might say.
But I see really with the Access to Justice movement and the way I became engaged with it is thinking about UPL as a barrier to, to access to justice, which I would argue that it is, despite very well meaning intentions behind our unauthorized practice of law rule. And I would say that today where we are is we have these two forces pushing on the unauthorized practice of law, and that is new types of legal service providers. And then of course, generative AI.
By the way, I'll just say no.
[00:03:19] Speaker B: 1.
[00:03:19] Speaker C: You know, for many years I did this work and UPL wasn't exactly a hot topic. So it is kind of fun for it to be a hot topic.
[00:03:27] Speaker B: I'm so glad to hear that. So that dovetails nicely into the next thing. I wanted to ask you about a relatively recent development.
So, last month, Nippon Life Insurance filed suit against OpenAI alleging, among other things, UPL after a litigant used ChatGPT to draft motions and tried to reopen and initiate lawsuits against Nippon.
So I'm curious and I'd love to hear your take on this. Were you surprised to see a UPL claim brought against one of these frontier AI companies, or did it feel like only a matter of time?
[00:04:02] Speaker C: That's a great. That's a very insightful question. I appreciate that. I think it was probably a matter of time. If we look back at the path of UPL cases, you know, when we, when we move beyond people giving legal advice, we started to look at things like forms. Could an individual.
Could Elizabeth draft, write, publish a book of forms that one might need to get help with their legal problem? Could Elizabeth publish that and sell it? And we decided a long time ago that yes, Elizabeth could do that.
And then we moved into software. And one of the first cases I remember was a case called Quicken Family Lawyer. And some of you may not be familiar with that, it's a Texas case, but Quicken Family Lawyer was the precursor to Legal Zoom. Before we had Legal Zoom, it was probably done on floppy disks, and it helped individuals determine what kind of legal assistance they might need in a family law case. And the state of Texas said this is the unauthorized practice of law.
Or the court in Texas said it was the unauthorized practice of law. And interestingly, very quickly thereafter, the Texas state legislature exempt such software from the prohibition on the practice of law by someone other than a lawyer. So we, we have created a family lawyer and then we got to Legal Doom and Legal Zoom. In the gosh, I guess early 2010s was a really big deal was the use of Legal Zoom, the unauthorized practice of law, based on how the software makes a determination. So if I log in or I go to Legal Zoom and I say I want to do maybe a will, and it asks me, do you have a spouse? And I pick yes or no, the decision that the software makes about what form I get, what Next question I get there was a big question about is this the practice of law? Because if it is, a non lawyer is doing it and we don't like that.
So we went from legal to other types of software. There was a case out of Florida called T I K D that was a, a, an app that would match someone who had a ticket with an attorney who could help them. But the app was not run by a lawyer. And the state of Florida, the, the Florida Supreme Court said this is the unauthorized practice of law.
So it seems very, very expected to me that we would eventually see a UPL action against the provider, the deployer of an open generative AI.
[00:06:57] Speaker B: Got it. And if I could ask a follow up question to that, it seems to me like a lot of those cases about legal tech companies, these enterprising companies that were providing access to justice survived in some way or another, whether it was by exemption or it was by adjusting what services they were providing to comply with the relevant UPL statutes. Do you see something like that happening here?
[00:07:20] Speaker C: Oh, that's a really, that's a great question because I think that is, I think that is possible. We're actually seeing that in Colorado right now in a, in a very limited way. But as you point out, you know, the Texas legislature took action. And then for example in North Carolina with Legal Zoom, we had a consent decree that said, okay, Legal Zoom, you can do this in North Carolina, but there's certain things that you have to do. And so we actually then created a legislative exemption. So we are seeing these exemptions. Colorado, I believe it was late last fall or middle of last fall, actually decided as a matter of fact, discipline from the State Bar, the licensing body for lawyers. They actually decided to do a pilot test of a non prosecution policy. This came out of there. They call it their Attorney Regulation Council, Office of Attorney Regulation Council. And they said we're going to adopt a non prosecution policy regarding unauthorized practice of law by non lawyers. Now this doesn't have the force of law, of course, but. And they require some consumer protection protection safeguards like oversight, some disclosures, some confidentiality protections, but they are essentially deprioritizing prosecution of developers of AIs. And so I see this as an example of just what you're talking about. Again, not the force of law and just a policy within their office. But this is the kind of thing that is we could very possibly see.
[00:09:02] Speaker B: So I wanted to ask, in the blog post that I wrote that accompanies this podcast episode, I explained why I think UPL rules as a normative matter should not be outlawing general purpose AI used by pro se litigants. And I explained why. I think the case for AI and access justice is quite strong, while the case for UPL and consumer protection, which is the traditional justification for UPL enforcement, is weak. And I mentioned that the Nippon vs OpenAI litigation is another example of a UPL complaint that did not originate from the consumer. So I'd love to hear from you. In your view, what is UPL enforcement in theory and in practice about?
And who are the typical drivers and players in UPL enforcement?
[00:09:51] Speaker C: Yeah, this is a great question and it's a really important question. Whether we're talking about new types of legal service providers or I should say human new types, types of human legal service providers or generative AI ads and whatever may follow that which we don't often talk about. Right. I think it's important to recognize that there's very well intentioned policy behind the regulation of the practice of law and that is bottom line consumer protection. We want to protect the public and we have created a lawyer licensing system with a goal of doing just that. You go to law school, you take an ethics test, you take certain classes, you graduate, there is a bar exam.
Right. You get sworn in, you comply with whatever licensing requirements that your state has. And all of that is based on. We need to do this to protect the consumer so that the consumer gets competent representation.
Now, when we couple that with the problem today, that is access to justice, most people can't afford access to civil legal services when they need help.
When we couple those two, I think it requires us to really step back and think about upl. And to your question, what is UPL about in theory and in practice or UPL enforcement about in theory and in practice? I think that the consumer protection justification is becoming weaker and weaker.
If you go back as far as some of the earliest UPL cases, there's a, there's a famous case that we, those of us who teach upl, teach about Mrs. Brumbaugh and she, she provided Maryland's secretarial services.
And the issue there was she was a non lawyer, she was not licensed to practice. And there are debates about the word non lawyer, but it's codified and I'm going to use it for simplicity's sake here.
It's not a great word, but Marilyn was not a lawyer. And there were questions about what she could do. She could type forms, but she could only put on the forms. These were legal forms.
What individuals directed her to write she couldn't change anything, she couldn't interpret anything. Right. All she could do was act as a scribe.
Okay.
And again, it was about consumer protection. This was an early case.
An action was brought against her.
Who brought the action?
The State bar, the lawyers in the state. And we see this over and over and over.
And I don't think that auto lawyers necessarily want these types of actions out of self protection. I do believe that many lawyers, and I think most that you would ask would say, I believe that we can only have consumer protection if it is lawyers who are delivering legal services.
But there is an increasing amount of data that suggests that that is not true. And when you couple that with the fact that we have a problem that people can't access legal services and we need it, I think it demands that we revisit things. And you point out the complaint did not originate from the consumer. The complaint did not originate from the users of Mrs. Brumbaugh's Maryland Secretarial service. The complaint in the Texas Quicken Lawyer case did not originate from consumers. It came from the Texas State bar. The legal zoom complaints did not originate from consumers. They came from state bars.
[00:13:51] Speaker B: Got it. That's really helpful. So shifting over to AI, sort of a big question I have, and I'm really curious to hear your thoughts on this.
In your view, what should be the relationship between pro se litigants and AI and why?
[00:14:06] Speaker C: Oh, this is a great question. This is the million dollar question. Right.
At the most fundamental, I believe that pro se litigants should be able to use the same tools that all who engage in the delivery of legal services can use. Can use.
Now is that realistic? No, they're not going to get access to the premium generative AI tools that are designed for use by lawyers. Why not? Because they're too expensive.
They're built on, you know, decades. Right. Of data. If you think about a Westlaw or a Lexis or these types of AI programs, and that involves cost and, and, and justifiably those companies need to recover some of those costs. Pro se clients or pro se litigants are not very likely to get access to those kinds of tools. But I don't think we can in good conscience prohibit pro se litigants from using free open source AI tools that anyone else could use.
That seems to me an absurd result.
So then the question that would be asked, and I'll tell you, I, I was at a wonderful conference last week and I want to compliment the national center for Access to justice and their work. They had an entire conference just a couple days ago on AI and access to justice and it was full of people providing legal services to those who may not otherwise have access to legal services and technology developers. It was really terrific. And, and, and when we had this conversation, you know, the question that then comes to me is so Ellen, if that's what you believe, how do we provide consumer protection? And that's a really hard question.
But the answer is not to prohibit pro se clients from using the tools themselves.
[00:16:11] Speaker B: And that aligns with something that I had written in my blog post, which was that I think it would be unfair to disallow pro se litigants from using the same tools that licensed attorneys and judges have access to, given that these litigants are the most disadvantaged participants in the courthouse. So thank you for that. And this, my next question relates to what you were just alluding to, which is that in my blog I also wrote about how I think other regulatory and disciplinary tools rather than UPL enforcement, are better suited to address real concerns about AI and consumer protection, so namely the risk of mistake and hallucination.
So I'm curious, in your view, how should we think about the trade off between, on the one hand, AI serving as a real resource for pro se litigants as we've spoken about, and on the other, AI enabling the rapid production of potentially mistaken filings. These hallucinations and errors have escaped detection at even well resourced law firms and are likely even more difficult for pro se litigants to spot.
[00:17:18] Speaker C: So what we need here is some form of balance. We need tools that work for consumers. We need consumers who are using tools that perhaps aren't designed for this, but for which they have easy access to, to understand what they're getting, what the tools are capable of.
That is all going to require a tremendous amount of education, which we aren't particularly great at and is a hard thing to do.
I think that there are a lot of organizations out there. I will, I will plug the North Carolina Legal Aid, which, and it is not the only one, but as a North Carolinian who are really working hard to create generative AI tools that can help those who can't otherwise access legal services and that do so based on an underlying data set that is appropriate for the questions that they're asking. So in other words, they're not going to an open AI, a chat GPT, a clawed pick your tool and asking a question to a database that is not tailored to answer the type of legal question that they have.
They are instead going to a tool that has been built, trained, vetted to provide information on, let's say, a problem that they might have with housing, a landlord, tenant issue.
These are, in my opinion, the ideal types of tools that individuals should be using.
It will not only provide a better resource for the pro se litigant who is using them, but it will if the problem is of a nature that it needs humans. And many problems are, let's be honest. Many people who need regal services not only may need a human, but they want a human. We can't forget that lawyers are not just attorneys, but we're attorneys and counselors at law. And a lot of people who fear AI worry about that counselor aspect. And I think that is a legitimate worry.
If we are deploying these tools through resources like legal aids that have the ability to provide human counseling to be able to know when a case is so unique, because there are cases like that, that it just doesn't fit the decision tree model and the generative AI may just not give the right answer, then those types of resources are available there.
The problem of course, is getting the word out about those resources. You know, I often think about my mom who is very good with the Internet and doesn't have a strong understanding of generative AI and might not know when she is, you know, using ChatGPT, exactly what she is using.
Disclaimers. You know, we give them a lot of credit in our legal system, but I'm not sure that they actually work. And in fact, in the, in the Nippon case, you know, chatgpt, and actually I'm not sure if the terms of service prohibited use for legal advice before the Nippon case was filed, but they certainly do now.
[00:21:07] Speaker B: So I'm curious also, you've written about the certification and regulation of new categories of legal service providers who might otherwise be considered unauthorized practitioners of law.
I'm curious if you might be able to share a little bit more about that work and do you think there are any lessons that are transferable to the area of AI and upl?
[00:21:30] Speaker C: Oh, I love this question, Elizabeth.
I think not only are they transferable, but I think that there is a good partnership that can be had here. And the background for those who may have focused on the AI world and not as much on the new types of legal services providers world is that there are because of the access to justice problem.
There are a number of states that are both entertaining and already have enabled new types of natural persons to deliver legal services in certain practice areas with certain training requirements. There may be educational requirements, Sometimes there's A certification test. Sometimes these individuals have to be supervised by lawyers. But nonetheless, we are seeing a boom in conversation and a pretty strong uptick beyond conversation in certification of these new types of providers.
These are being utilized despite unauthorized practice of law prohibitions that would otherwise prohibit their work.
Because what we've recognized is that this well meaning policy of consumer protection has, in my opinion, resulted in over regulation and an assumption that the way in which we license lawyers today, this current framework, ensures lawyer competence.
Now, this licensing process that we have, and you're going to know this firsthand, Elizabeth, as my students do, it comes at a great cost.
A great cost. And not just in dollars, broadly defined, a great cost. And so then the question becomes, right, do we really need all of these levels right. In our licensing system to provide competent representation? So states are saying we're willing to experiment in the same way that Colorado is saying we're willing to have a policy of not prosecuting, you know, generative AI deployers who are providing legal services.
States are starting to experiment here and I'm very pleased with this experimentation. I think that perhaps we're going to see more and more experimentation.
Lawyers don't necessarily love this, right? It's the lawyer monopoly, which has fought others engaging in the practice of law all the way back to Mrs. Brumbaugh's Secretary of Service through Quicken Lawyer, through LegalZoom, and now through Generative AI.
They some well meaning, some lawyers very well meaning and saying, look, this is about consumer protection and others saying, you know, this is my life's work and this is my bread and butter and if all of a sudden someone can do it who doesn't have a law degree, this is really going to be disruptive and really put me out of business.
So it's a monopoly though. And not only are we a monopoly, we're a self regulated monopoly.
And we have the opportunity I think right now to rethink how we regulate that monopoly to increase access to justice to individuals. And it's hard. It's very hard. It's a hard thing to ask of lawyers. I think, though, to your question directly, if you combine generative AI and new types of legal service providers who have some training, who understand the legal issues that folks are facing, but may not be, quote, licensed lawyers and give them these generative AI tools that work well, combine the two, allow folks who can't otherwise access legal services to have the benefit of both a generative AI designed to help them get information to solve their legal problems, and a human to help them with their counseling need. I think there's real opportunity here.
[00:25:59] Speaker B: Thank you so much. I have two follow up questions to that, if that's okay. The first is for our listeners. Could you just give some examples of these new categories of legal service providers?
[00:26:09] Speaker C: Absolutely. So what we are seeing is very, very popular right now is called a community justice worker. Community justice workers started in Alaska. These are individuals who are volunteers. They are not getting paid for the delivery of legal services. They are placed within areas where. And by areas, I mean they are placed in like a medical facility or other areas where an individual who comes in with a problem may have a legal problem but may not even know it.
They are typically supervised by the legal aid. That's exact, that's how it works in Alaska. There are new programs popping up. I think there's, gosh, I don't know, Elizabeth. There may be eight, nine or ten. I can, I can get you some data on that if you Google Frontline Justice. They are on the, on the front lines literally of the community justice worker movement.
So we have community justice workers. We have housing advocates in Delaware.
I'm not sure of their exact name, but that's the area in which they work with landlord tenant issues. Because there's some extraordinary data about how so many tenants are unrepresented when they have a housing issue with their landlord. In South Carolina we have a group of folks who we do call housing advocates who are working under the supervision of the NAACP in South Carolina and they are providing legal services.
Arizona has a group they call legal paraprofessionals. They can work in certain family law matters, some administrative work, juvenile work.
Minnesota also has legal paraprofessionals. Colorado has the same.
Utah and Oregon have various versions of it. So and then we have some like navigators and document preparers who I would say are not really engaged in the practice of law. But just to be safe, we are creating or states are creating certifications for them so that they don't cross that line.
[00:28:30] Speaker B: And my second follow up question was you had mentioned that the legal bar constitutes a sort of monopoly. And I'm curious, how viable do you think if an AI provider were to make an antitrust theory against state bars for monopolizing legal service provision. I know that LegalZoom had made that argument against the North Carolina bar. And I'm wondering what you think of that as a potential route.
[00:28:56] Speaker C: That's such a great argument, Elizabeth. And here's what I'm going to say about that.
So you have an economics master's degree. If I had an economics master's degree. I would have been an antitrust lawyer.
I don't know that I have. I don't know that I have the math chops to do antitrust law.
What I will say about that, and this is based on my very limited knowledge about antitrust law, is that it hasn't worked yet.
I won't make a prediction about the future.
[00:29:28] Speaker B: Got it. Thank you so much.
[00:29:29] Speaker C: But it's a great question.
[00:29:31] Speaker B: I'm curious to see how it will play out, if it does.
[00:29:34] Speaker C: Absolutely. Okay.
[00:29:36] Speaker B: I've also thought about the uneasy fit between UPL rules and forms of legal assistance, such as legal aid, which you've talked about, and also jailhouse lawyering, which largely fill gaps where we otherwise don't have services in place. And I'm just curious if you have any thoughts on the role of UPL enforcement in those spaces.
[00:29:57] Speaker C: Oh, gosh, yes. I.
I don't think that we have done a great job differentiating UPL enforcement.
Differentiating UPL enforcement when there's no one else to assist. Right. In other words, where there's a true access to justice need and where there's not.
And I think that one of the things that I'm seeing when you bring up legal aid, and this is this is. This may be slightly off your question topic.
One of the things that I'm seeing that troubles me a little bit and when you said delayed it, it brings it up, is some of these new programs that I just talked about, the qualified tenant advocate in Delaware and community justice workers, they require these new types of providers to be supervised in perpetuity.
And that troubles me because what that does is that's going to shift a burden and create a resource problem for the lawyers who are assigned to supervise them. Now, supervision is a good thing. And there may be a period time where supervision is perfect almost in an apprenticeship like model or while the new provider is learning what it is that the new provider needs to know. But the perpetual supervision, or a lack thereof resulting in a violation of the UPL restrictions is very troubling to me.
And I don't know that much about UPL enforcement in jailhouse lawyering, but I know that a lot of good has come from what you refer to as jailhouse lawyering.
[00:31:49] Speaker B: Got it. I'm also curious. So in UPL rules, there's the classic distinction between legal information and legal advice. And I'm curious, when it comes to AI, do you think that that line between information and advice is still meaningful?
[00:32:05] Speaker C: I don't, Elizabeth, and I have really wrestled with this. I would actually be curious for your fault. I don't think it's meaningful because, you know, if we're talking about an old school software, a Quicken or a legal zoom where we have a decision tree and everything that comes out is based on, you know, it's extracting based on something that was used to build the software, but you're not getting anything new that's clearly, you know, legal. Legal information in my opinion. But I just, I'm just not convinced yet that we can go so far as to say that a generative AI, just because it generates something new, is actually providing legal advice. Now interestingly, in the Nippon case, you know, if you read that complaint, which I have several times, I mean the allegations are things like GPT was a legal assistant and advisor, it intentionally induced and facilitated, it encouraged and assisted in the motion. So it's really, you know, giving those human attributes to the algorithm. I'm just not there yet.
[00:33:22] Speaker B: That's really interesting. Yes, I think it's difficult because I mean these AI systems purport to be helpful, honest and harmless. Right. So helpful is one of their main principles. And to be helpful to be that assistant, the line between information and advice becomes really hazy.
So that's, it's helpful to hear your perspective. I appreciate that.
[00:33:41] Speaker C: I agree, I agree.
[00:33:43] Speaker B: So in my blog post I ended with the prediction that the toothpaste is already out of the tube. Sort of like what happened with Uber. So given the widespread use of AI today, it may be just unrealistic to meaningfully enforce UPL rules against it now. So I would love to hear, if you had to guess, what do you think the future of AI and UPL will look like and what do you think it should look like?
[00:34:08] Speaker C: Oh gosh, this is a great and a hard question, so I agree with you. The toothpaste is out, we're not going back. These are now, you know, thoroughly woven into our society as a whole.
Folks are and by folks, you know, states and others are trying to get a handle on things. I think we are behind in governance and at the same time I think some of the governments that is being put in place is not that thoughtful. New York has proposed legislation that would make tool that engage in any type of medicine, law, social work, nursing, etc, would really impose civil liability on the proprietor of the tool which is, you know, that will be, that'll be an interesting determination. But if the software produces any advice, so any medical advice, any social work advice, any legal advice, but it doesn't carve out nonprofits or groups that are trying to create, develop, provide tools that are based on good underlying data that can really meaningful, meaningfully assist with access to justice.
So I certainly don't think it should look like that. I do think we need to be moving into talking about good governance and what that might look like, talking about good education for folks and what that might look like. And if we are going to hold groups like OpenAI accountable, if we're going to hold them accountable, you know, it can't be under a UPL framework. You know, maybe it's negligence, I don't know. But UPL doesn't fit there.
[00:36:05] Speaker B: So those were all the questions that I was hoping to ask you, which has been a really wonderful conversation, and I just wanted to give you the space. Now, if there's anything else you'd like to add that I haven't asked you about.
[00:36:15] Speaker C: Oh, I really appreciate that. This is fantastic. I'm so excited about your interest in this. And I would just say that, you know, for those of you listening, there's a lot going on in this conversation. There are a number of folks who are talking about this, working in this area. I encourage you to look at the work at the national center for Access to Justice. I encourage you to look at the work of Satish Nori, who has done a lot of work in this area.
They are really providing thoughtful information about how we should be thinking about these systems.
And I would encourage those of you who have never tried a system to just try it. And that's probably not many of your listeners, Elizabeth, but it does continue to amaze me how many people want to talk to me about AI who have never tried it, because I think that's so important whenever, whatever tool you try just to sit down, whatever tool you decide just to sit down and try it. And I, and I want to say I think there are also a lot of people who are very thoughtfully engaged in how do we provide proper governance. There's a group AI now, who's doing some thoughtful work there, and I think we need more of that as well. But I do think we all have to stop talking. And we have more. We should continue talking, but we need to start implementing governance policies simultaneously because this is happening fast.
[00:37:49] Speaker B: Well, thank you so, so much, Professor Murphy, for taking the time to speak with me today. I really enjoyed our conversation and I'm really grateful to have the chance to speak with you and to hear your thoughts.
[00:37:59] Speaker C: Oh, thank you so much, Elizabeth.
[00:38:01] Speaker A: Proof over precedent is a production of the access to Justice Lab at Harvard Law School.
Views expressed in student podcasts are not necessarily those of the A J Lab.
Thanks for listening. If we piqued your interest, please subscribe wherever you get your podcasts. Even better, leave us a rating or share an episode with a friend or on social media.
Here's a sneak preview of what we'll bring you next week.
[00:38:27] Speaker D: More than 20 jurisdictions have recently passed Right to Council policies that provide free lawyers to tenants facing eviction.
And you know, these policies have remarkably little evidence, and it's not at all obvious that such policies would work for reasons that we can get into. I kind of came across an opportunity where I was working with implementation partners on that other project, studying what happens when you give people emergency rental assistance, where they received funding to provide lawyers to tenants facing eviction, where the idea was we want to offer people more robust opportunities to defend themselves in court. Providing cash assistance on its own may be ineffective without somebody to kind of shepherd the money to the landlord. And so they wanted to evaluate the program. And so based on our existing partnership, I was able to persuade them that it would be really valuable to try to set this up in a clean, randomized, controlled trial.
