chevron-down Created with Sketch Beta.

Law Practice Magazine

The Big Ideas Issue

Lowering UPL Risk for Nonlawyer Legal Help

Lucian T Pera


  • The rise of legal AI tools is expanding access to affordable legal services, allowing consumers to obtain customized legal documents and legal advice at a lower cost than traditional lawyers.
  • To mitigate UPL risks, providers of AI legal tools should distinguish between legal information and legal advice, use clear disclaimers, educate users about the tool's limitations and ensure data privacy.
Lowering UPL Risk for Nonlawyer Legal Help

Jump to:

The shadow legal services market is growing in all directions.

As the bright new day of legal artificial intelligence (AI) dawns—or lawyers’ long professional nightmare of legal AI descends, your choice—the number, scope and utility of AI- and software-driven tools available to anyone with a legal problem is quietly exploding.

Those lawyers and others who want to join this market need to think seriously about the risk of being found to violate the law barring the unauthorized practice of law (UPL). There are concrete steps they can take to mitigate those risks. We’ll explore those in this column.

The Emerging World of Consumer Legal Help

Today, ordinary people can get wills and contracts customized to their personal needs drafted for a fraction of what human lawyers charge.

Services can draft or help people fill out court forms to defend themselves in a collection or divorce case. Other services will review a legal document—say, a lease or employment agreement—and provide a detailed evaluation of its terms and its legal meaning and effect.

Yet other consumer tools can answer legal questions in plain English—What sentence is my son facing for his DUI arrest last night? Can my landlord refuse to return my security deposit when I move out?

And these services can be cheap or free—in any case, they cost less than a human lawyer.

Legal Tech, We Have a Problem

But there’s a problem. Several, actually.

First, of course, these services and tools have to get it right.

Some big brains—human and digital—are cracking that one, as you read this. Some lawyers and public-interest legal organizations, including legal aid groups, are working hard to develop and road test services like this to get the answers right and be as user-friendly as a cheerful and empathetic legal aid lawyer.

So are for-profit businesses, some owned and run by lawyers, some not.

I predict the competence problem will be addressed directly, and AI tools will rapidly become competent at “simple” legal problems and move on to more complex and individualized problems. I expect this will happen sooner than most lawyers believe.

Do You Want to Go to Jail?

Second, bigger problem: all of this could be illegal, and even criminal.

For a hundred years or so, every U.S. jurisdiction has had laws and court rules on the books that say that only licensed lawyers can engage in the “practice of law.” Then they define the scope of what that covered as the “practice of law” in broad, sweeping terms.

For example, one formulation defines the practice of law as “the application of legal principles and judgment with regard to the circumstances or objectives of a person that require the knowledge and skill of a person trained in the law.”

Yes, that is a bit circular. Welcome to the law governing UPL.

Local Variations

Your mileage will vary. UPL laws varies significantly among the jurisdictions. The sweeping breadth of some states’ laws is narrowed a bit in other jurisdictions. And every jurisdiction has all kinds of statutory, case law, and rules-based exceptions. “Patchwork” is a polite way to describe the crazy quilt of UPL law.

Indeed, several public-interest groups have launched challenges to various forms of UPL prohibitions in New York, South Carolina and North Carolina, on First Amendment grounds, but these each involve efforts by nonlawyer humans to provide legal help in the face of UPL prohibitions.

AI and UPL

So, riddle me this unauthorized practice bar exam question.

Suppose an AI tool is trained—successfully—to learn, understand and explain everything about landlord and tenant law and eviction. The maker of this AI embeds it in an app that sells its services for $25 per consultation. The app is heavily advertised outside the county courthouse. And let’s assume the app asks for and analyzes the user’s specific lease terms, payment history and current economic situation, and then provides the user a script for what to file and say in court. Maybe it even provides the user with a paper or pleading to file on their behalf in court. Is that UPL?

Well, if you follow the mainstream approach of most UPL laws, this app is giving legal advice to the user about their specific factual situation—their lease, their payment history, maybe even the landlord’s failure to fix the plumbing. Worse yet, it’s preparing them to represent themselves in court—wait, that’s what lawyers do!

Few UPL laws require that a human be involved. Very few consider whether the activity is done for compensation.

And to be clear, this is not a new problem. Connecticut and New York both prosecuted Norman F. Dacey for UPL for simply publishing his book, How to Avoid Probate!, essentially a forms book that guided readers through forming trusts. (Spoiler alert: Dacey lost one and won one.) And Parsons Technology Inc. lost its UPL battle in Texas courts over its publication of Quicken Family Lawyer—just selling the will-writing software was held to be UPL—until the Texas legislature amended Texas’ UPL statute to specifically say software-driven tools were not UPL. That law remains on the books.

Still, for decades now, there hasn't been much meaningful UPL enforcement of people selling legal help in form books. Nobody has dared to try shutting down TurboTax.

Enter AI

That brings us to the problem. As the bright new day of legal AI dawns, innovators are trying to use technology—software, bots, AI and other tools—to bring legal help to consumers who need it and are often willing to pay for it. But they are faced with the yawning uncertainty of whether what they are doing is illegal, whether they will get sued or whether they may even be prosecuted as criminals for doing so.

In this author’s opinion, shared by many who work in the field, UPL is one of the two biggest impediments to innovation in legal services—the other being the ban on fee sharing with nonlawyers.

So let’s assume you are an innovator—lawyer or not—exploring the use of AI to deliver legal help to consumers without lawyer involvement. You know that UPL enforcement could be a risk to your service. How can you mitigate your risk in providing an AI- or software-driven service that provides legal help?

Mitigate Your Risk

Legal information versus legal advice. There is a well-established distinction between providing legal information—even legal information on a narrow and well-defined topic—and providing legal advice. The former can be done by anyone and should not be considered UPL. The latter, especially when directed to the specifically articulated problems of a user, generally constitutes UPL. Quite clearly, the more an AI tool provides advice or a service specifically aimed at a user’s specific situation or needs, the greater the risk of a UPL finding.

This is not, however, the end of the inquiry. Under the law of many jurisdictions, the frontier lays at the point where an AI tool, well-programmed and trained by a lawyer, can evaluate a common legal document—like a residential lease, an NDA or an employment agreement—to look for the presence or absence of specific prohibited or recommended terms, or to explain the meaning of other terms in plain English. For centuries, this has been the province of lawyers and considered the practice of law. But AI tools can today perform this function.

In my view, the reader of a lawyer-written article that lays out the required, prohibited or preferred terms needed in a type of document—like some that have appeared in this space—does not commit UPL by reading that article and applying its teachings to their own document. If rather than write an article, I train an AI tool to do this for a user who then uses the AI tool without my knowledge or help, I do not believe that should be considered UPL. Of course, that may not be the law, yet or ever.

Other Protections

Disclaimers. Be very clear in disclosing to all users—and have each agree as a part of terms of service—that the provider of the services is not a lawyer or law firm, that the services provided are not a substitute for legal services from a lawyer,; that the user may consult a lawyer for advice, that using the service does not establish a lawyer-client relationship and that no lawyer-client confidentiality or privilege is established by use of the services.

Be sure to say all this in plain, clear, conspicuous English. Repetition is the motion of learning, and it may be helpful to include disclaimers along these lines at key points in the user’s customer experience, as well as on any documents provided the user. Remember, too, that even if you have a user agree that you are not providing them legal advice or legal services, if you do, in fact, provide them legal advice or legal services, you may yet be found to have engaged in UPL. Disclaimers are not a cure-all.

Disclose and educate about your limitations. Depending on the application, it is also probably not enough to simply make the type of disclosures set out above. It’s in the interest of any service provider for their user to understand fully both the reach and limitations of the service. Perhaps the service provider should provide detailed educational information about the area in which the service is provided. Perhaps the provider should explain in serious detail what the service does and does not do. Setting realistic user expectations, apart from generally enhancing the customer experience, certainly reduces the likelihood that the user will believe they are receiving legal services when something else is intended and delivered.

Confidentiality. Even if you effectively disclaim any confidentiality obligation, it is conceivable that some of the information your service obtains from users may be subject to protection under data privacy laws. It is well worth your time and money to seek the advice of a data privacy lawyer on these issues. Moreover, simply to gain and maintain the trust of consumers, you may well need to make certain commitments to privacy or confidentiality.

Legal and ethics review. Strongly consider consulting with a lawyer trained in UPL law—maybe an ethics lawyer—to help you assess the UPL risks, as you design, before you launch and as you iterate and change the tool. Your tool will doubtless change; the way users use it will change; and the law in this area is dynamic.

Despite the dire warnings above about UPL’s breadth, there are jurisdictions where they may be some protection for some software-only activity. As mentioned above, Texas has such a law, and other jurisdictions have some decisional law that might be protective. Get advice from a lawyer on this up front.

Moreover, in some jurisdictions, engagement with UPL regulators may be possible and even productive. Only a local legal expert, knowledgeable in local UPL law and enforcement, will be able to advise on this subject. In many jurisdictions, regulatory advice is either simply unavailable or inadvisable. Still, there is no substitute for local expert guidance in evaluating the risk environment.

A New Age Dawns?

Perhaps The Golden Age of Legal Self-Help is dawning. But if you are working with AI tools to deliver legal help, be careful out there.