Arizona Law Journal of Emerging Technologies
Volume 7 Article 3, 02-2024
Download Article Here
Image from PEXELS.COM

ARTIFICIAL LAWYERING: A JEKYLL AND HYDE STORY

Nicholas R. Spagnuolo*

I. Abstract

Artificial intelligence (AI) is increasingly used by the public to make important decisions, from the advancement of business algorithms to aiding with college admissions processes. Much has been written recently about AI and the law. But where is the intersection between AI and the law? And where is the line drawn to separate the two? This Article addresses the benefits and detriments to the use of AI in the legal setting, and provides a solution through the Model Rules of Professional Conduct to capture the benefits while simultaneously abating the detriments.

The discussion aims to break down the typical black-box description by computer scientists who create their own lexicography into an understandable, easily discernable analogy. To that measure, I first discuss AI generally to create a baseline understanding. I then turn to how AI is used in the current legal setting by both lawyers and laypersons. A key motivation to writing this Article is to highlight the systemic problems plaguing the legal system, and how AI can resolve these problems. However, science has taught us that every action has an equal and opposite reaction. The benefits of legal aid through AI are contrasted with discussions about how the use of AI in the legal field can create more issues than it solves, such as implicating the unauthorized practice of law (“UPL”), endangering the public, and magnifying the effects of inherent bias in data sets. Finally, this Article suggests a proactive solution to reap the benefits of AI and minimize the potential problems that arise. Instead of an overview of AI and the law, this Article explores the ethical implications and policy arguments surrounding the use of AI in legal settings. In light of AI transforming the legal industry in numerous ways every day, society—and more specifically lawyers—must maintain their ethical principles in the changing environment.

II. Introduction

ChatGPT and other forms of generative AI have the ability to revolutionize the legal industry or destroy it. One of the biggest advances in technology is Artificial Intelligence (“AI”).¹ Recently, we have seen a never-ending list of news articles being published daily about AI, from wacky tales² to stealing human jobs.³ AI is now capable of creating sentences, paragraphs, and responses that are indistinguishable to that of a human.⁴

As the development of generative AI has led to machines replicating the speech of a human,⁵ it is not hard to envision an artificial human on the other side of the webpage answering your questions. People experience this phenomenon all the time with chatbots on insurance websites,⁶ banking websites,⁷ and even while shopping online.⁸ The modern trend of using chat functions over phone calls has grown in popularity. OpenAI’s ChatGPT is one of the most advanced artificial intelligence algorithms, which frequently answers the users’ questions accurately and efficiently.⁹ This proverbial artificial human on the other side of the chat appears to know the answers to every question and allows people to get these answers without the nuisance of having to make small talk about the weather. As faith in the artificial human grows, many may very likely wonder, what else it can do? How else can this super intelligent being help? Does this artificial person have any limits?

When the artificial human is asked questions, it provides answers. If the questions are legal, then the AI becomes more specialized and transforms into what can be most aptly described as an artificial lawyer. The artificial lawyer, who we will call “AL,” will appear to be super helpful and will attempt to answer all the questions the user can conjure. However, AL has two sides, a split personality. This split personality is analogous to an old gothic novella from the late 1800s, The Strange Case of Dr. Jekyll and Mr. Hyde.¹⁰ The tale involves one personality, a Dr. Henry Jekyll, who is a reliable gentleman with many friends and is esteemed in society.¹¹ The other personality, a Mr. Edward Hyde, is a criminal, murderer, and detriment to society.¹² The artificial human in our case, AL, has the same split.

On one hand, AL has the capability of being Dr. Jekyll. By day, the revolution of AI can provide massive benefits to attorneys as the next great tool used to accomplish more work in a shorter period of time and increase client satisfaction. AL can provide legal assistance to low-income individuals who the law seems to ignore.¹³ Generative AI can benefit non-lawyers by providing information and access to justice, helping them not only determine whether they have a legal issue, but also how to tackle the issue directly.¹⁴ Moreover, AL can increase the public perception of the legal system, which would be met with a surge of trust, leaving society more confident and empowered.¹⁵

On the other, AL could be Mr. Hyde. By night, the revolution of AI can cause large amounts of concerns by violating legal and ethical rules. Thus, AL would be a criminal and could face the full wrath of the legal system for violating the law.¹⁶ Of course, as an AI algorithm, AL is not a licensed attorney and would be partaking in the unauthorized practice of law. Thus, artificial intelligence is capable of injuring society with no redressability for the damage done. This would hurt public perception and trust in the legal system. As trust in the system is already low,¹⁷ the injuries caused by AL can lead to a dystopia and vast amounts of damage to American citizens. And if you are a lawyer, you are probably thinking about how these machines can take your job from you, creating a dystopia of legal unemployment.

The duality of the reputed artificial lawyer is currently at the fingertips of internet users and must be examined. This Article explores both the ethical and the unethical implications of artificial intelligence in the legal setting and argues for an update to the Model Rules of Professional Conduct to reflect the change in technology. The use of generative AI can help to resolve the access to justice issues in America; however, conversely, the use of AI for legal aid can go too far and violate unauthorized practice of law statutes. Part III exposes the problem of how the advancement of technology intersects with the legal profession by exploring the concept of artificial intelligence and giving a brief overview of the ethical rules against the unauthorized practice of law (“UPL”). Part IV discusses the benefits of using artificial intelligence to provide legal recourse to a group of persons who believe the courtroom is not an option for them. Part V examines the UPL implications of artificial intelligence participating in the practice of law. Finally, Part VI discusses how the Model Rules of Professional Conduct could be amended and how the legal community can reap the benefits of generative AI. Each benefit and drawback must be considered when attempting to resolve the issues presented in artificial lawyering. Artificial lawyering requires balancing access to justice, ethical considerations, and technological drawbacks associated with AI models.

III. The First Step of Problem-Solving is Understanding the Problem: A Concise Introduction to Artificial Intelligence and the Rules of Lawyering

a. Artificial Intelligence Technology

AI has been shown to outperform humans and provide amazing results to complex problems.¹⁸ The benefits expedite public dissemination. Society is drawn to the new and exciting, and to ideas that help accomplish a goal in the most efficient manner. The powers of AI are both remarkable and worrisome. Yet, artificial intelligence has a long and winding history and did not always have the same definition it does now.¹⁹ AI has changed over time and grown to have many different meanings.

i. What is AI? How does it work?

Although AI has been around for over sixty years, the majority of people still do not understand how it works.²⁰ The “black box” of AI is mysterious to the users of the algorithms and can even be a mystery to the creators.²¹ AI is defined as “a system’s ability to interpret external data correctly, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation.”²² However, AI is an abnormal scientific pursuit with a constantly changing definition in order to match technological growth.²³ To aid with narrowing scope, this paper will only focus on understanding the inner workings of generative AI. This type of AI, such as ChatGPT, is capable of generating human-like answers to a user’s query and answer complex questions,²⁴ which implicates AI lawyering the most.

Machine learning is a broad area of research focusing on making computers that develop algorithms, which can improve themselves through experience.²⁵ Machine learning has emerged as the ideal method for subareas of AI research, such as “computer vision, speech recognition, natural language processing, robot control, and other applications.”²⁶ ChatGPT is powered by “a class of machine learning Natural Language Processing models known as Large Language Models (LLMs).”²⁷ In general, tools utilized by machine learning systems initially are trained by large numbers of data sets linked in some way to help words associate with each other.²⁸ How artificial networks use the relationships between words can be analogized to a common party card game: “Apples to Apples.”

In case you are unfamiliar with the rules of Apples to Apples, the game generally begins when a judge places a “green apple card” on the table and then the other players choose a “red apple card” from their hand which best represents the word on the green card.²⁹ Thus, the word on the green apple card has an intrinsic value the players are trying to match with their red cards, which also contains a word or group of words with intrinsic value. The judge then goes through and “selects the [red card] he or she thinks is best described by the word on the green apple card.”³⁰ In order to keep score, the winning duo of cards is kept and logged.³¹ For example, the green card may have the word “slippery.” Each player puts in a red card that may have “ice skating,” “glue,” and “taking a bath” on them. Clearly the intrinsic value of the word “glue” does not match the value of “slippery,” and thus, it would be eliminated from contention. Then it is a judgment of whether “ice skating” or “taking a bath” more closely resembles the judge’s preference when the judge thinks of the word “slippery.” Either card could win as both are close enough in value. As the game continues, the players start to understand the judge better and play cards according to the judge’s preferences and can increase their chance of winning.

How would this pertain to machine learning and natural language processing systems? The “judge” is the input user for the AI system and the “green apple card” is the query presented by the user to the AI model.³² The AI model represents the players, with different data sets representing the different hands of “red apple cards.”³³ Just like the words have value on each of the cards in the game, the AI has given value to words in its system.³⁴ Then, just like the game, the AI system/players presents answers to the AI user/judge from their data set most closely resembling the value of the query.³⁵ The AI system does this for each word or word set, stringing together a sentence to sound like natural language.³⁶ This represents the natural language processing system.³⁷ For systems like ChatGPT—a LLM—the models incorporate a large amount of text data to infer relationships between the words and the texts.³⁸

Relative to machine learning, the winning duo of cards are logged, and the losing set is also noted.³⁹ The AI system—just like the players—can look back and analyze what it did right and wrong to better match the judge’s preferences. The training and ability to match the user’s preferences is similar to how AI learns from itself through machine learning.

ii. Current Legal AI

The battle between AI-powered legal services and traditional legal services has been a part of the legal community for more than a decade.⁴⁰ LegalZoom is one example of AI-powered legal services that generates standardized legal documents based on user inputs.⁴¹ LegalZoom differs from the complex model of ChatGPT, as the program inserts data on forms already made and reviews answers for consistency and completeness.⁴² The software utilizes questionnaires created by customers to generate legal documents.⁴³ These documents are made through an “automated process” which is “generated from standardized language” based on the inputs from the customer.⁴⁴ Thus, LegalZoom is a low-level AI system and differs greatly from complex AI programs like ChatGPT.⁴⁵

The program found itself in legal battles based on its alleged unauthorized practice of law.⁴⁶ However, LegalZoom appeared to get the proverbial “green light” from at least one higher state court in 2014, as the South Carolina Supreme Court adopted the findings of Circuit Judge Clifton Newman, stating the program did not violate state law.⁴⁷ Judge Newman used the general business model of LegalZoom to determine that the preparation of legal documents without providing “advice, consultation, explanation, or recommendations on matters of law,” is not practicing law and LegalZoom did not violate UPL.⁴⁸ Yet, with the development of artificial intelligence, not all of AI can fit into the LegalZoom classification.

Recently, ChatGPT has emerged as a new and exciting artificial intelligence system and can respond to inquiries like a human might. With new developments made in the computer science field, new questions arise. How smart is ChatGPT? Can ChatGPT become smarter and more aware than the average human? Is ChatGPT crossing into the practice of law? Although the first two questions involve deeper considerations into philosophy, this Article critically tackles the last question to get clarification on the use of artificial intelligence in the practice of law.

The combination of programs can create a newly operated artificial lawyer capable of providing legal descriptions from ChatGPT for the legal documents made through LegalZoom. The ability for ChatGPT to aid in legal work almost instantaneously emerges with the current version of the program.⁴⁹ There is the potential for companies and firms to use the program to draft legal documents and have their own set of forms,⁵⁰ similar to how LegalZoom has a set of forms in their repertoire. As technology advances, so does the potential for AI programs to penetrate the practice of law. Of course, each program must answer to unauthorized practice of law claims. Lawyers will end up on either side of the debate depending on if they see a net benefit or detriment to society.⁵¹ Some companies—such as DoNotPay, the “self-proclaimed World’s First Robot Lawyer”—have already utilized “GPT[] technology to help users negotiate down bills and fight traffic tickets.”⁵² With ChatGPT passing law school exams⁵³ and the bar exam,⁵⁴ these questions need to be answered sooner rather than later.

b. Model Rules for Lawyering in America

i. Legal Ethics

Courts have joined with bar associations to regulate lawyers through self-regulation and set the standards for rules and ethics in the profession.⁵⁵ The self-regulation of lawyers dictate the process of admitting lawyers to practice and punishes any unprofessional behavior committed by lawyers.⁵⁶ Professional bar associations wanted “greater control over the practice of law;”⁵⁷ and thus, the American Bar Association created the Model Rules of Professional Conduct (“Model Rules”) to provide a framework for regulating the conduct of lawyers and guide lawyers to act ethically.⁵⁸

The Model Rules were adopted by states to reform lawyer disciplinary processes and were seen as an improvement to existing law.⁵⁹ Nonetheless, it has been 50 years since the Model Rules were adopted by the ABA House of Delegates in 1983.⁶⁰ Although the rules have been amended “in the context of advances in technology and global legal practice developments” during a Commission on Ethics 20/20 review,⁶¹ neither yielded any resolution or report for the growing field of AI.⁶² Recommendations from the Commission focused on advancements of technology for malware, multijurisdictional practices, and a lawyer’s use of internet based client development tools.⁶³ The Commission last proposed resolutions based on technology developments on February 11, 2013.⁶⁴ With the growth of technology and its associated ethical concerns, another ABA Commission on Ethics may be necessary.

From the drafting of the rules, each state has passed their own set of laws pertaining to the ethical conduct of lawyers.⁶⁵ However, the ABA Model Rules and revisions to the rules are often adopted as the ethical rules and standards in each state.⁶⁶ With almost every state and the District of Columbia requiring the Multistate Professional Responsibility Examination based on the Model Rules,⁶⁷ a change to the rules could have a nation-wide ripple effect within each state and avoid conflicting jurisdictional treatment of AI.

ii. The Unauthorized Practice of Law

Other rules of ethics may be implicated in the growth of interest and use of artificial intelligence in the legal field.⁶⁸ However, the rule most connected with artificial lawyering is the unauthorized practice of law (“UPL”), which was raised when LegalZoom started to grow in popularity and continues to be a major concern in legal practice.⁶⁹ Rule 5.5 of the ABA Model Rules sets forth the limitation of practicing law without a license:

(a) A lawyer shall not practice law in a jurisdiction in violation of the regulation of the legal profession in that jurisdiction, or assist another in doing so.
(b) A lawyer who is not admitted to practice in this jurisdiction shall not:

(1) except as authorized by these Rules or other law, establish an office or other systematic and continuous presence in this jurisdiction for the practice of law; or
(2) hold out to the public or otherwise represent that the lawyer is admitted to practice law in this jurisdiction.⁷⁰

The Model Rules are considered to only apply to lawyers, but most—if not all—jurisdictions have a rule stating a person who is not authorized to practice law may not engage in the unauthorized practice of law.⁷¹

The comments to the rule highlights that each jurisdiction typically has different localized definitions for the “practice of law.”⁷² The definitions are vast and ambiguous,⁷³ but some states have statutes that hunt non-lawyers who are providing any legal services.⁷⁴ Other states have developed the understanding of unauthorized practice of law through common law.⁷⁵

The entire nation has grappled with a workable definition of the practice of law.⁷⁶ The Supreme Court of South Carolina seems to voice a nationwide consensus, where the best course of action for defining the “practice of law” is for the highest court “to decide what is and what is not the unauthorized practice of law in the context of an actual case or controversy.”⁷⁷ Some examples of what has been held as the practice of law in the past include drafting legal documents,⁷⁸ offering specific legal advice,⁷⁹ and putting in appearances at court for a client.⁸⁰ Furthermore, the Supreme Court of South Carolina held LegalZoom was not contributing to the practice of law when the service faced scrutiny.⁸¹ Yet, the case against LegalZoom was closer in other states and came to settlements or arbitrations with state bar associations.⁸²

Thus, the practice of law does not have a clear definition, but a state court will allegedly “know it when they see it” and weigh the benefits and drawbacks when ruling. The UPL implications of AI services beyond LegalZoom likely need to be evaluated due to growth in other technology. Ultimately, testing the UPL implications of ChatGPT in a case or controversy would be a novel issue for courts to address.

IV. Artificial Lawyers Would Aid in Unattended Areas of the Law

The benefit of allowing artificial intelligence to aid individuals is to provide an answer to a problem far too common in America. The legal system was meant to protect every citizen, to protect the rights of people who cannot protect themselves, and to ensure everyone is treated equally under the law.⁸³ Yet, it seems the idealized words of Thomas Jefferson to have a “perfect Union”⁸⁴ and to “establish Justice”⁸⁵ has fallen short of its mark.

The American legal system has an immense access to justice issue.⁸⁶ Low-income members of society are unable to afford legal counsel, which intensifies systemic inequalities and disadvantages present in society.⁸⁷ Some groups of interest receive little or no help from the legal system, including veterans, persons with disabilities, parents of children under the age of eighteen, and survivors of domestic violence or sexual assault.⁸⁸

Mary Hicks is just one example of a low-income individual who faced a common legal issue.⁸⁹ For two years, Mary paid $975 per month for a Washington D.C., apartment with mold and mildew in the bathroom and holes in her wall.⁹⁰ Through the help from a legal clinic, she was able to avoid eviction, get repairs to her apartment, and discover her unit is rent controlled.⁹¹ However, everyone is not as lucky as Mary.⁹² Eighty-six percent of the civil legal problems faced by low-income Americans are done without efficient legal counsel.⁹³ Additionally, people who experience an overabundance of civil legal problems only seek professional help around twenty percent of the time.⁹⁴ The burden is on the legal system to ensure Jefferson’s words are not empty promises, but rather goals to strive for while using what is available at the time. This goal includes providing an avenue for people to access justice.

Access to justice can be defined as “the improved delivery of legal services, to all citizens, by enhancing accessibility and constantly improving efficiency.”⁹⁵ There are a surplus of legal technologies that can be used, and are used, in order to accomplish the goal of delivering legal services to all citizens.⁹⁶ The advancements in technology do not need to be revolutionary to help the legal system prosper.⁹⁷ However, revolutionary technology can expedite the availability of a solution. Artificial intelligence is a revolutionary technology which has the capability to alter the cultural perspective of the legal system and open the doors of the courtroom to every citizen.

A portion of access to justice is to provide information to citizens about legal recourse for their issues. One of the main reasons low-income Americans do not seek professional legal help is because they are not sure whether the problem they are facing is “legal in nature.”⁹⁸ When people do not seek professional help, they turn to other resources, such as asking friends and family for advice or searching for the information online.⁹⁹ AI can be the chief source for legal knowledge. Unlike search engines, which require navigating different websites to parse through opinions and statutory language, generative AI is able to instantly give an answer that appears to be written by a human.¹⁰⁰ Additionally, the large amount of data allows for the AI model to synthesize different points into one concise answer.¹⁰¹ When confused, the person putting an inquiry into the algorithm could ask for definitions and clarification.¹⁰² This allows for a single question-and-answer prompt for individuals to explore and learn about legal information.

The use of AI to help educate the public about legal issues and increase public awareness on common justiciable issues allows for individuals without access to justice to seek legal help. For example, a low-income individual—such as Mary Hicks—could ask ChatGPT, “can I sue my landlord for the mold and mildew in by bathrooms and having holes in my wall?”¹⁰³ From this, ChatGPT would provide the following response:¹⁰⁴

The response provides a user with their options, disclaims the advice is simply for education, and allows for further inquiry.¹⁰⁵ The response also allows an individual to learn new legal vocabulary such as “breach of the lease agreement” and “negligence,” which can be explored further to help with the civil legal problems faced by low-income Americans.¹⁰⁶

The importance of educating low-income individuals on their legal recourse and options is amplified if they intend to represent themselves pro se. A majority of individuals believe they can successfully represent themselves in court as pro se litigants and trust the court will support them in obtaining justice.¹⁰⁷ Yet, the expectation for the court to help is wrongly placed, as judges are required to uphold impartiality in both fact and appearance.¹⁰⁸ Pro se assistance policies and programs are developed due to the “awareness of the significance of the loss of public trust and confidence in the courts to our orderly society.”¹⁰⁹

Forty-two states have adopted some recommendations to help self-represented litigants and ten states have adopted comprehensive delivery systems to allow access to legal self-help.¹¹⁰ The challenge is to “ensure every person in those states have access to some level of legal self-help.”¹¹¹ Artificial intelligence can be seen as a somewhat pro se assistance that is used to educate on specific state laws that would be available to a society at large. Moreover, other pro se litigant guides help to describe the filing process (and even how to dress),¹¹² but they do not aid with any substantive law or descriptions.¹¹³ To get help in understanding whether or not their issue is “legal” and how to understand specific laws, a person would have to do a separate research or click through link after link to get the information needed to properly represent themselves.¹¹⁴ The creation of a centralized location for a pro se litigant to get all the information necessary, and in a way they can understand, will allow for pro se litigants to understand the elements and tests required to be proven and met, respectively. Additionally, this material is already available to pro se litigants online. Unlike suggestions for court staff to provide this information to the litigants,¹¹⁵ the use of AI is a 24/7 endeavor with the capability of answering large amounts of users at once. The fairness of proceedings for all litigants will increase the public faith in the legal system and allow for each and every citizen to access the law meaningfully.

At this point, perhaps you are wondering “the situation cannot be this dire, can it?” Or “there are certainly enough lawyers to handle these cases if they just focus a little more on pro bono.” However, after considering some basic facts, it is apparent the legal profession will only benefit from the use of AI.

According to the American Bar Association, there are over 1.3 million lawyers in the United States.¹¹⁶ Even with more than a million lawyers, low-income individuals “will receive insufficient or no legal help for an estimated 1.1 million eligible problems.”¹¹⁷ As observed, the majority of poor folk do not receive any help.¹¹⁸

Why is this? The bulk of lawyers work in private or corporate legal offices who represent corporations and institutions, not low-income individuals.¹¹⁹ The “drift” from helping low-income persons starts in law school.¹²⁰ When students enter into law school about half “express a desire to represent underserved clients or causes,” but “the proportion . . . after graduation falls to less than 3 percent.”¹²¹ Students considering whether to go into private industry or the public sector often express skepticism about the law’s ability to actually produce social change and see the private sector as more significant because of the opportunity to make a significant income.¹²² With the disinterest of the legal community to represent those in need, “the number of lawyers who regularly represent poor people in civil cases is about 6,000.”¹²³ Less than 1% of lawyers in the United States help low-income Americans.¹²⁴

Because lawyers are not representing the area of law plaguing low-income individuals, the use of AI is not taking away from the legal profession, but rather adding the opportunity of legal recourse to those who seek answers. Using technology as a legal tool can increase access to justice with ease because it is both workable and replicable.¹²⁵ The increased access to justice will allow pro se litigants to feel more confident, more empowered, and can change the outcome of the case.¹²⁶ Artificial intelligence is a technological tool with legal implications “that brings us closer to fair resolution” and would have “resounding implications on the justice system.”¹²⁷ The rule of law cannot substitute for social capital and trust.¹²⁸ Trust in the legal system is very important in society as a whole.¹²⁹ The trust gained by opening the doors of the legal system to individuals who have been turned away more often than not would foster cooperation between informal and formal mechanisms of social cooperation.¹³⁰ Therefore, AI will not only aid people who are ignored by the law, but it will have incidental benefits to the legal profession by increasing societal trust in the system.

Another concern with using AI to educate individuals on the law is the potential for an increase of frivolous lawsuits clogging the dockets of state courts. Many believe there is an excessive number of frivolous lawsuits in the civil justice system.¹³¹ State courts already have unwieldly dockets filled with countless cases, some of which may be frivolous.¹³² Meanwhile, litigation is slow, and justice could be delayed to individuals who are seeking an answer through the courts.¹³³ Yet, the estimation of frivolous lawsuits is overblown and the belief Americans are too litigious can be considered a myth.¹³⁴

Lawyers have a duty to say “no” to a lawsuit they believe will be frivolous.¹³⁵ This is enforced through the Federal Rules of Civil Procedure.¹³⁶ Although this duty is also applied to pro se litigants, many courts will exhibit patience to assure them the same access to justice.¹³⁷ Still, these litigants may not recognize when their suit is frivolous. Some have contracted their way out of their day in court, and others are unable to sue due to immunizations to wrongdoers.¹³⁸ There could be an influx of unknowingly frivolous lawsuits in the court system, which would crowd the dockets even more. This not only hurts the court system, but it would slow down the justice sought by all litigants, whether they be pro se litigants or attorneys.¹³⁹ The existence of such suits hurt the trust in the legal profession.¹⁴⁰ As with any new revolutionary idea, the risk and reward must be considered, and new technologies bring new issues to the forefront.¹⁴¹

Overall, society would benefit by allowing AI to be a legal tool to indigent persons, opening doors to the courthouse that have been previously closed off. There are also concerns about the system disfavoring pro se litigants, a loss of opportunities for attorneys, and an influx of frivolous lawsuits. Many of the concerns are inflated by unsupported fears plaguing the civil justice system. Although not a perfect solution, artificial intelligence can help an area of law ignored by lawyers for decades. The only question remaining is whether AI goes too far and is partaking in the unauthorized practice of law; and thus, is violating the law.

V. Artificial Lawyering is Illegal Lawyering

There are major concerns both in the generative AI models, with AI biases and unforeseen secondary problems;¹⁴² and legally, such as with citing non-existent legal precedent,¹⁴³ which require reeling in the lofty expectations placed on systems like ChatGPT. One such concern is when a generative AI model goes beyond being a tool by participating in the unauthorized practice of law, and hurts a user with no legal redressability.

a. Is ChatGPT Capable of the “Practice of Law?”

ChatGPT is capable of the practice of law. Although the practice of law is difficult to define and most courts have taken a ‘we’ll know it when we see it’ approach to this area of the law,¹⁴⁴ some constant UPL rulings through different jurisdictions show ChatGPT is practicing law. The “practice of law” is generally understood to be representing a person in proceedings before a court of justice.¹⁴⁵ The definition is much more expansive and includes legal advice, preparing legal instruments and contracts, and is independent of whether a lawsuit has been filed.¹⁴⁶ The inquiry does not depend on what can be “measured by the comprehension of a trained legal mind, however, but by the understanding thereof which is possessed by a reasonably intelligent layman who is reasonably familiar with similar transactions.”¹⁴⁷

In one example, the Supreme Court of Ohio determined “that the practice of law encompasses the drafting and preparation of pleadings filed in the court[] . . . and the preparation of legal documents and instruments upon which legal rights are secured or advanced.”¹⁴⁸ Similarly, in Missouri, it was determined preparing documents that are legal in nature can be considered the practice of law¹⁴⁹ as the legislature there has defined “practice of law” to be “the drawing of papers, pleadings or documents . . . in connection with proceedings pending or prospective before any court of record. . . .”¹⁵⁰

Many federal courts have taken the practice of law a step forward when analyzing Rule 11 violations¹⁵¹ and denounce the custom commonly known as “ghostwriting.”¹⁵² Similarly, the Rhode Island Supreme Court has determined the preparation or drafting of legal documents on behalf of a client “unequivocally constitutes the practice of law.”¹⁵³ The court prohibits attorneys from helping pro se litigants with:

[T]he preparation of pleadings, motions, or other written submissions unless the attorney signs the document and discloses thereon his or her identity and the nature and extent of the assistance that he or she is providing to the tribunal and to all parties to the litigation. The attorney shall also indicate on the written document, if applicable, that his or her signature does not constitute an entry of appearance.¹⁵⁴

Thus, the act of preparing legal documentation in order to aid pro se litigants involves practicing law.

The question now turns to whether ChatGPT is capable of any of the above actions, as ChatGPT would violate UPL rules if it contributes to providing legal advice or preparing documents for pro se litigants. ChatGPT is capable of informing a user whether they have a legal issue on their hand.¹⁵⁵ However, if a user goes one step forward and asks ChatGPT to “Write a complaint for a breach of lease agreement in Washington, D.C.,”¹⁵⁶ the chatbot obliges:¹⁵⁷

ChatGPT’s ability to write complaints would qualify as unethical lawyering. Complaints are the beginning stages of a lawsuit and a legal instrument where rights are secured, which would satisfy statutory definitions.¹⁵⁸ On the other hand, whether the legal document is fully “prepared” would be a fact-specific inquiry state courts would have to determine. Whether to draw the line at the drafting of legal arguments or inputting facts would be a better question for state courts, or, more likely, state legislatures. Drafting complaints is the practice of law.¹⁵⁹ Courts have determined filling in blanks on forms is not the practice of law.¹⁶⁰ Therefore, if the user is not practicing law by filling in blanks, then drafting the remainder of the pleading would meet the requirements of illegal lawyering.

ChatGPT is not limited to pleadings. The generative AI program is capable of engaging in the practice of law in regarding other legal documents, such as responding to an Office Action from the United States Patent and Trademark Office (“USPTO”). In LegalForce RAPC Worldwide, P.C. v. DeMassa,¹⁶¹ the Northern District of California determined the plaintiff failed to show the defendant’s employees engaged in the practice of law when responding to USPTO Office Actions.¹⁶² The key fact included that if an applicant’s trademark was rejected through an Office Action, then the applicant has the opportunity to challenge the determination.¹⁶³ To challenge the determination, some trademark applicants have referred the matter to Trademark Express, an entity owned by DeMassa, who was not an attorney.¹⁶⁴ Trademark Express assisted customers by helping to “handle” an Office Action.¹⁶⁵ If a customer of Trademark Express wanted to challenge a rejection based on “likelihood of confusion” or contest the USPTO’s determination the attempted trademark is generic, then Trademark Express would defer to an attorney.¹⁶⁶ Thus, the defendant believed drafting a response to the Office Action would qualify as the practice of law, and prohibited itself from engaging in responses to the USPTO that would start to resemble the practice of law.¹⁶⁷ The court seemingly agreed.¹⁶⁸ While humans can make the judgment call, ChatGPT cannot. When prompted to write a “response to the Office Action that says my trademark is generic, using case law,”¹⁶⁹ ChatGPT offered this response:¹⁷⁰

The Northern District of California would likely see this as crossing the line into violating state UPL laws. If the action done by a human of layman intelligence would constitute illegal lawyering, then the same standard should be used for a program of artificial layman intelligence. Therefore, it is clear that ChatGPT is capable of—and does engage in—the practice of law.

b. Allowing ChatGPT to Practice Law Without Authorization Will Endanger the Public

Early legislation against nonlawyers practicing law focused on three evils: (1) the practice of “mercenary attorneys” (i.e., attorneys seeking to stir up litigation), (2) incompetence of untrained professionals, and (3) the charging of excessive fees for the services performed.¹⁷¹ Similar evils could arise again if there is a determination that ChatGPT is not violating UPL rules. “Mercenary” legal documentation preparation companies could arise and attempt to provide the same acts at ChatGPT, increasing the number of nonlegal entities creating “fill in the blank” documents and stirring up additional litigation.¹⁷² The competence of the training of the AI model would have to be evaluated more meticulously to verify competence and protect against biases in the system confounding from the legal system’s past.¹⁷³ Furthermore, it is not hard to envision a future where individuals who provide tangential aid to overstep into the practice of law, and challenge UPL restrictions based upon the allowance of AI to participate in the same types of document preparation and legal advice. Opening the door to AI participating in the practice of law will restart the long journey taken to arrive at current UPL rules.¹⁷⁴

The purpose of regulating the unauthorized practice of law is to “protect the public against incompetence, divided loyalties, and other attendant evils that are often associated with unskilled representation.”¹⁷⁵ The purpose is not to maintain a monopoly on the industry, as some critics would suggest, but to avoid a legal system filled with “unqualified persons over whom the judicial department can exercise little, if any, control in the matter of infraction of the code of conduct which, in the public interest, lawyers are bound to observe.”¹⁷⁶ Attorneys have the responsibility to comply with the law and to understand the extent of both legal and professional rules.¹⁷⁷ Moreover, duties are placed on lawyers to maintain certain protections for clients, such as the duty of competence,¹⁷⁸ duty of confidentiality,¹⁷⁹ and the duty charge reasonable fees.¹⁸⁰ The same duties are not implicated on individuals who fall outside the reach of legal self-regulation. The public protection goal of UPL also implicates the protection from violations of the related duties presented in the Model Rules.¹⁸¹

Not everyone agrees with the claimed purpose of unauthorized practice statutes.¹⁸² Critics believe the main goal of UPL legislation is to maintain the lawyer’s monopoly on providing legal services.¹⁸³ Lawyers are not popular, and by sustaining a stronghold on legal practice, consumers are injured through the inability to have a free market.¹⁸⁴

Monopoly of the legal system has also led to the substantial access to justice issue in America.¹⁸⁵ Lawyers determine if they are willing to take a case, which leaves plaintiffs with modest potential recoveries or limited potential to pay fees, without counsel.¹⁸⁶ A monopoly through UPL statutes prevents anyone from representing modest plaintiffs.¹⁸⁷ Critics believe only true accomplishment of UPL statutes is limiting voluntary transactions between individuals, leaving some markets ignored and allows lawyers to increase legal service fees without competition to counteract.¹⁸⁸ A free market can resolve a large majority of the concerns used in defense of UPL statutes, as individuals are unlikely to “ask a nurse to perform a heart operation, a bookkeeper to perform a difficult accounting analysis, or a patent lawyer to defend against a murder charge even if doing so appeared cheaper than the alternatives.”¹⁸⁹

However, the free-market argument has major flaws. Indeed, it is unlikely an individual would ask a nurse to perform heart surgery, but what if it is difficult to tell the nurse and the doctor apart? What if the nurse has performed the surgery once before and the patient did not die? What if the patient did die and no one knows? Instead of having a multitude of deaths—or in this case legal injuries—on the conscious of the regulatory scheme, state bar associations have acted proactively to protect the public much faster than a market could react. UPL statutes protect legal consumers. Meanwhile, AI lawyering is in a free market. Although consumers can tell they are speaking with AI chatbots instead of an actual human, many believe the chatbots are capable of accomplishing their tasks more efficiently and lower in cost.¹⁹⁰ As public trust in AI grows, the potential for injury to plaintiffs using AI for their legal issues grows exponentially faster, as one AI program can serve many clients at once. UPL statutes are not perfect, but the balance between protecting the public and limiting lawyers’ monopoly should favor overprotection.

The fact remains that unauthorized lawyering can injure the public. If a human lawyer violates a duty that is placed on them, then they could be liable for malpractice claims from the client hurt by the lawyer’s actions.¹⁹¹ If an individual is hurt by a mistake made by ChatGPT, then there is no redressability for that injury.¹⁹² Civil liability of lawyers ensures key goals of the legal system are met, such as satisfying fiduciary duties, mitigating overly adversarial behavior harming the public, and increasing accessibility to legal services.¹⁹³ Holding a lawyer liable through legal malpractice depends on reasonableness and practice norms.¹⁹⁴ Courts will attempt to use the same standard with artificial lawyers. The soundness of a court attempting to mirror the reasonableness standard with the one for a human lawyer, where courts “balance the cost of precaution against the risk averted multiplied by its probability . . .”¹⁹⁵ is murky at best. Moreover, there is no practice norms to compare to the actions of artificial lawyers. Scholars have suggested how legal malpractice should be defined: “[F]ailure to provide a client the services that a lawyer of ordinary knowledge, skill and diligence reasonably should provide.”¹⁹⁶ This raises questions of whether an AI algorithm could provide the same level of diligence or knowledge to drafting legal documents. It is likely that AI alone will commit legal malpractice.

Additionally, if a lawyer violates duties recognized by their state bar association, then the lawyer can be liable for different sanctions and enforcement strategies to deter the action in the future.¹⁹⁷ Other than civil liability to the client, other enforcement strategies include “criminal prosecutions, . . . disbarment or other ethical sanction, contempt of court, [and] sanctions in the current legal proceedings . . . .”¹⁹⁸ Disbarment and ethical sanctions are the most commonly used strategy to deter any breach of an ethical duty.¹⁹⁹ Bar associations are specially situated to determine if a violation has occurred and to design an appropriate punishment.²⁰⁰ Yet, critics believe the resolutions to deter violations of ethical rules are slow, expensive, and favor lawyers due to inherent biases.²⁰¹ Society has an interest in protecting the public through compliance with applicable professional rules.²⁰² Whether an extension of UPL is the correct solution depends on the importance society places on protection when compared to potential injurious conduct by technology operating as a lawyer.

Bar associations also lag behind the realities of current practice.²⁰³ Current and future law graduates will face a new world that legal professionals were incapable of envisioning ten to fifteen years ago. To that end, the threat of liability may deter individuals from participating in legitimate uses of generative AI to educate, as well as the illegitimate conduct of practicing law.²⁰⁴ The deterrence effect could go too far and be a significant cost to both generative AI companies and users of the algorithms.²⁰⁵

Yet, the fact remains that clients rely on the legal system to provide them recourse for damages they incur, which includes the damages incurred when lawyers commit malpractice and the deterrence of improper actions by lawyers.²⁰⁶ The improper actions of a generative AI lawyer have no clear path to the courtroom. There is no law to hold an AI program liable, and questions arise as to who an injured plaintiff could sue. Do you sue the computer scientist who created the algorithm? The company who owns rights to the algorithm copyright? Or is the injured plaintiff required to suffer in silence? The legal system is not prepared to answer these questions, and the legislative body is not primed either. Other nations have suggested ways of holding AI liable outside of the legal malpractice context,²⁰⁷ but these recommendations face scrutiny for failing to properly acknowledge the social and ethical challenges related to accountability.²⁰⁸ The answer may come from each state legislature or the federal legislature, but the legal system can, and should, proactively act within its own self-regulatory scheme.

Overall, allowing ChatGPT to practice law could crowd courts with frivolous lawsuits, raise concerns about AI not following legal ethics, and prevent individuals injured through AI malpractice to receive redress.

c. Additional Concerns Facing ChatGPT Lawyering

In addition to the legal issues with ChatGPT participating in the practice of law, there are technological challenges AI programs face. These challenges have the potential to injure any user in an AI legal setting, whether the use is legitimate or illegitimate.

One such challenge is the limitation of language and how not all language is representative in the text of data sets.²⁰⁹ A language model may be able to speak, or write, but “that doesn’t mean it understands what it is talking about.”²¹⁰ Humans are able to understand sentences based on deeper understandings of the contexts that surround the sentence.²¹¹ Yet, AI is limited in its understanding; the ability to use language is very different from the ability to explain.²¹² In the context of artificial lawyering, the ability to use language is not sufficient in aiding individuals with their legal issues. A sophisticated layperson using the program may be able to satisfy the last logical leap, but generative AI will be available to everyone, not just the sophisticated. The language limitation can cause damage to an AI user who is searching for information, and the damage would be increased if the user is searching for something more.

Another challenge facing generative AI is the inherent bias present in data sets.²¹³ Bias in machine learning includes the “assumptions made by a specific model.”²¹⁴ AI relies on data generated by humans, so any bias existing in human society is projected onto the AI model.²¹⁵ The model is then capable of reproducing, or increasing, the bias present in any particular society—causing a larger disadvantage to minority groups that are discriminated against.²¹⁶ The American legal system already has discrimination and bias woven into its existence.²¹⁷ Human actors within the system have potential biases reaching across every area of law.²¹⁸ An artificial lawyer will lack the understanding of the bias and reproduce the bias independent of intentions of the input user. The justice system should fight against biases and discrimination, not reward and intensify them.

VI. A Proactive Solution to Artificial Lawyering

A solution to the problems of AI lawyering should focus on a balance between the benefits of access to justice and the dangers associated with violating the rules of professional conduct. Other areas of the Model Rules implicated the same balancing test and allow the client to waive the apparent violation in order to capture the benefit.²¹⁹ To cure the associated dangers, these areas of the rules require the attorney to obtain informed consent from the client in order to proceed.²²⁰ Similarly, the Model Rules should be amended with a comment to Rule 5.5 to (1) recognize the growth of technology and how the unauthorized practice of law is implicated and (2) establish the benefits of unauthorized practice of law by technology through the informed consent of clients.

a. Update to the Model Rules of Professional Conduct

Similar to updates made to Rule 5.5 of the Model Rules in response to changes in technology for multijurisdictional practices, the American Bar Association should update the comments associated with the rule to account for new technological advances in artificial intelligence.

Currently, Rule 5.5 states: “A lawyer shall not practice law in a jurisdiction in violation of the regulation of the legal profession in that jurisdiction, or assist another in doing so.”²²¹ The comments explain the practice of law is established by law and varies jurisdictionally.²²² A proposed change would be the addition of a comment to establish that technology is capable of the practice of law and how to cure this deficit. The suggested language for the comment could be:

[22] The use of a practicing entity to practice law is prohibited in all jurisdictions. This Rule does not limit the use of practicing entities for services equivalent to the services of paraprofessionals and delegates, so long as the lawyer supervises the delegated work and retains responsibility for its work. The definition of a practicing entity is established by law and varies from one jurisdiction to another. However, in general, a practicing entity is any non-human entity capable of satisfying the practice of law requirement in each jurisdiction. A practicing entity may provide services to a pro se litigant and resolve the problem under this Rule through obtaining informed consent, confirmed in writing. Pro se litigants are required to disclose to the court their use of a practicing entity.²²³

This solution allows for the use of technology, while balancing the benefits and drawbacks preemptively. The solution defers the definition of a practicing entity to the different jurisdictions, while concurrently establishing the entity is anything capable of performing legal actions. The additional comment recognizes the use of technology, and the subsequent growth of technology by having a broad definition of “any non-human entity.” Technology will not be hindered, but the use of technology for law will be regulated by ethical rules.

Most importantly, the update establishes that technology is capable of the unauthorized practice of law if the model ventures into any definition established by a jurisdiction. By explicitly showing artificial intelligence can violate UPL statutes, the rules of professional conduct will deter the use of AI models to practice law. Critics may argue the deterrence effect is too great and will limit uses of AI as a tool for legal information.²²⁴ Yet, these actions would not rise to the level of a UPL violation, as it is simply informing. The issues that rise to the level of artificial lawyering include drafting legal documents for a layperson and providing legal advice.²²⁵ Bar associations will be protecting society from injuries caused by an unqualified AI system attempting to satisfy an input user’s query.

In addition to protecting society, the comment adds for a limited resolution to the Rule for pro se litigants, who are the individuals most benefited by artificial lawyering. The cure of the violation will help close the justice gap that is plaguing the United States.²²⁶ By requiring informed consent in writing, both the AI company and users are protected from liability issues. Informed consent is important to maximize benefits to lawyers, clients, and society in general.²²⁷ The rules of consent are used to protect clients from misunderstandings in the future and to protect the individuals who owe a duty from the same miscommunications.²²⁸ Also, informed consent allows for autonomy and trust in the system, especially for indigent clients.²²⁹ By allowing indigent individuals and pro se plaintiffs to access justice through informed consent, they are protected from any misunderstandings of the consequences they may face later in litigation from using generative AI. Additionally, the generative AI companies are protected from the liability associated with UPL statutes.

Therefore, a comment should be added to Model Rule 5.5 based on the new technological advances in artificial intelligence in order to capture the potential benefits of AI while concurrently protecting society from detrimental effects.

b. Rule 11 Changes the Classification of Generative AI for Lawyers

The addition of a comment would benefit laypersons and AI companies from the dangers presented through artificial lawyering. However, it is important to note lawyers are held to different standards and the use of generative AI does not present the same dangers.

Lawyers are required to follow rules of professional conduct or face liability and sanctions in response to any violation.²³⁰ Moreover, the Federal Rules of Civil Procedure require lawyers participating to sign “[e]very pleading, written motion, and other paper,” and to avoid delaying the court or making frivolous arguments.²³¹ Therefore, lawyers using generative AI have a duty and responsibility to review the work as if the AI was being done by a paraprofessional, such as a law clerk.²³² The attorneys who delegate this work to an AI system retain the responsibility that comes from any violations.²³³

Thus, the use of generative AI to lawyers implicates that the AI is not a practicing entity, but rather a tool. This double standard presents some issues, such as the addition to the monopoly lawyers have on legal industry.

VII. Conclusion

Artificial intelligence, specifically generative AI, has an enormous amount of potential to disrupt the world with positive change. Along with the potential to benefit the world, if used incorrectly, AI can cause injury to many people. One instance where the use of a generative AI system can help or harm individuals is through artificial lawyering—the Dr. Jekyll and Mr. Hyde of artificial intelligence.

On one hand, the use of generative AI can help to resolve the access to justice issues in America. Through education of legal principles to help pro se litigants prepare for what to expect in a courtroom, AI has the potential to be a powerful tool for an area of law the majority of lawyers ignore. Low-income individuals will be able to obtain legal solutions for their problems and will help to facilitate a new era of trust.

On the other hand, the use of AI for legal aid can go too far and violate unauthorized practice of law statutes. Unauthorized practice of law statutes are passed to protect the public from improper legal techniques and any injury that flows therefrom. Artificial intelligence has technological issues with a lack of understanding and bias which may hurt a user for which there is no redressability. Legal duties are not implicated on individuals who fall outside the reach of legal self-regulation. Society has an interest in protecting the public through professional rules, and the same professional rules should be used to protect the public from artificial lawyering harm.

Therefore, in order to reap the benefits of sophisticated AI algorithms capable of practicing law, the American Bar Association should add a comment to the Model Rules. Within this comment, it should clarify the issues around whether technology is capable of UPL and prohibit its use in order to protect society. However, the prohibition cannot go too far as to deter companies from growing their technological capabilities and prevent low-income individuals from seeking aid for their legal issues. The violation of UPL by technology can be resolved through informed consent, which benefits all parties involved. Therefore, the comment balances two competing interests: (1) the growth of technology and everything that comes along, and (2) the benefits of the use of AI technology can be captured through informed consent.

* J.D. Candidate, University of South Carolina Joseph F. Rice School of Law (expected May 2024); B.S. Biomedical Engineering, University of South Carolina (2020). I would like to thank Professor Bryant Walker Smith for guiding my exploration into this complex legal issue and challenging me to think critically about the law. Also, special thanks to Professor Laura Lane-Steele for her insight and feedback, without which this Article could not have come to fruition.

Footnotes

1
Brady D. Lund et al., ChatGPT and a New Academic Reality: Artificial Intelligence-Written Research Papers and the Ethics of Large Language Models in Scholarly Publishing, 74 J. Ass’n Info. Sci. Tech. 570, 571 (2023).
2
See 12+ Funniest ChatGPT Conversations, Capitalize My Title (Dec. 26, 2022), https://capitalizemytitle.com/funny-conversations-with-chatgpt/.
3
See Aaron Mok & Jacob Zinkula, ChatGPT May be Coming for Our Jobs. Here Are the 10 Roles that AI is Most Likely to Replace, Bus. Insider (Apr. 9, 2023, 11:36 AM), https://www.businessinsider.com/chatgpt-jobs-at-risk-replacement-artificial-intelligence-ai-labor-trends-2023-02.
4
See Lund, supra note 1, at 572.
5
See Jacob Browning & Yann Lecun, AI and the Limits of Language, Noema Mag. (Aug. 23, 2022), https://www.noemamag.com/ai-and-the-limits-of-language/.
6
Chatbots, Nat’l Ass’n of Ins. Comm’rs, https://content.naic.org/cipr-topics/chatbots (last updated Apr. 3, 2023).
7
Ron Shevlin, Digital Banking Didn’t Kill Bank Branches—But Chatbots Will, Forbes (Nov. 14, 2022, 9:00 AM), https://www.forbes.com/sites/ronshevlin/2022/11/14/digital-banking-didnt-kill-bank-branches-but-chatbots-will/?sh=3b27443e2c7f.
8
Over Half of Online Shoppers Who Use Chatbots Are More Likely to Shop from Brands Using Conversational AI, Yahoo Fin. (Apr. 10, 2023, 8:00 AM), https://finance.yahoo.com/news/over-half-online-shoppers-chatbots-120000897.
9
Anil Ananthaswamy, In AI, Is Bigger Always Better?, Nature, https://www.nature.com/articles/d41586-023-00641-w (last updated Mar. 10, 2023) (“Artificial-intelligence systems that can churn out fluent text, such as OpenAI’s ChatGPT, are the newest darlings of the technology industry.”).
10
See Robert Louis Stevenson, The Strange Case of Dr. Jekyll & Mr. Hyde (The Floating Press 2008) (1886).
11
See id.
12
See id.
13
John Villasenor, How AI Will Revolutionize the Practice of Law, Brookings (Mar. 20, 2023), https://www.brookings.edu/articles/how-ai-will-revolutionize-the-practice-of-law/.
14
Opportunities & Risks for AI, Legal Help, and Access to Justice, Just. Innovation (June 28, 2023), https://justiceinnovation.law.stanford.edu/opportunities-risks-for-ai-legal-help-and-access-to-justice/.
15
See Villasenor, supra note 13.
16
See infra note 74 (examples of laws prohibiting the unauthorized practice of law with criminal sanctions).
17
See Benjamin H. Barton, An Institutional Analysis of Lawyer Regulation: Who Should Control Lawyer Regulation—Courts, Legislatures, or the Market?, 37 Ga. L. Re. 1167, 1167 (2003).
18
See Weiyu Wang & Keng Siau, Artificial Intelligence, Machine Learning, Automation, Robotics, Future of Work, and Future of Humanity: A Review and Research Agenda, 30 J. of Database Mgmt. 61, 63–66 (2019) (describing the potential of AI in business, healthcare, education, military, and other areas impacting society).
19
See Rockwell Anyoha, The History of Artificial Intelligence, Sci. in the News Harv. Univ. (Aug. 28, 2017), https://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/.
20
See Roger C. Schank, What is AI, Anyway?, 8 AI Mag. 59, 59 (1987) (“[A]rtificial intelligence is almost completely misunderstood by individuals outside the field.”).
21
See Wang & Siau, supra note 18, at 67.
22
Michael Haenlein & Andreas Kaplan, A Brief History of Artificial Intelligence: On the Past, Present, and Future of Artificial Intelligence, 61 Cal. Mgmt. Rev. 5, 5 (2019).
23
See Schank, supra note 20, at 62.
24
Owen Hughes, Generative AI Defined: How It Works, Benefits and Dangers, TechRepublic (Nov. 21, 2023), https://www.techrepublic.com/article/what-is-generative-ai/ (discussing different examples of generative AI and how “generative AI models are increasingly being incorporated into online tools and chatbots that allow users to type questions or instructions into an input field, upon which the AI model will generate a human-like response.”).
25
Michael I. Jordan & Tom M. Mitchell, Machine Learning: Trends, Perspectives, and Prospects, 349 Sci. Mag. 255, 255 (2015).
26
Id. (emphasis added).
27
Molly Ruby, How ChatGPT Works: The Model Behind the Bot, Towards Data Sci. (Jan. 30, 2023), https://towardsdatascience.com/how-chatgpt-works-the-models-behind-the-bot-1ce5fca96286.
28
See Bryant Walker Smith, Dall-E Does Palsgraf, 14 Case W. Res. J. L. Tech. & Internet 89, 91 (2023) (Dall-E is a comparable AI system to ChatGPT which uses visual and text elements instead of text-to-text elements to create an artificial neural network).
29
Apples to Apples Game Rules, Ultra Bd. Games, https://www.ultraboardgames.com/apples-to-apples/game-rules.php (last visited Mar. 31, 2023).
30
Id.
31
Id.
32
Many algorithms focus on problems with input-output functions and improving the function to optimize the performance metric. See Jordan & Mitchell, supra note 25, at 255–56. A core method widely accepted by AI systems involves supervised learning methods, “where the training data take the form of a collection of (x, y) pairs and the goal is to produce a prediction y* in response to a query x*.” Id. at 257. Another core method is reinforcement learning, where “[i]nstead of training examples that indicate the correct output for a given input, the training data . . . are assumed to provide only an indication as to whether an action is correct or not.” Id. at 258. Thus, the overall goal a modern artificial intelligence system is to receive an input, reason from the input, and explain an output with human-like interactions to provide decision support for specific tasks. Artificial Intelligence: What It Is and Why It Matters, SAS, https://www.sas.com/en_us/insights/analytics/what-is-artificial-intelligence.html (last visited Apr. 1, 2023).
33
Data sets are used to train the model in text classification – “the process of categorizing texts . . . into organized groups.” Shervin Minaee et al., Deep Learning-based Text Classification: A Comprehensive Review, 54 ACM Computing Survs. 62:1, 62:3 (2021). Text classification involves both recognizing text that appears in the input and generating answers on the fly. See id. Within the analogy, the red apples are the data that is stored within the system in order to accomplish the generative portion of the input/output process.
34
AI models used to “integerize” word types, where each word was given a nonnegative value and every word was stored and indexed. See Noah A. Smith, Contextual Word Representations: Putting Words Into Computers, 63 Commc’ns ACM 66, 68 (2020). The assignments were arbitrary, alphabetical, or in the order the system read them, but it gave a value to the words that could quickly test whether two integers, and thus, two words, were identical. Id. Natural language processing models have now started to use vectors to describe words and tokens. Id. A vector is a mathematical term to describe an object that has both magnitude and direction; and, if one thinks of an arrow on a graph, the magnitude is the length of the arrow, and the direction is which way the arrow is pointing. See Jerrold E. Marsden & Anthony Tromba, Vector Calculus 4–6 (6th ed. 2012). Each word is now represented as a real-valued vector and the “distance or angle between pairs of word vectors [is] . . . the primary method for evaluating the intrinsic quality of such a set of word representations.” Jeffrey Pennington et al., GloVe: Global Vectors for Word Representation, ACL Anthology (Oct. 24, 2014), https://aclanthology.org/D14-1162.pdf. For purposes of this Article, it is only important to understand that AI models give values to words and those values are used to help the AI system understand the relation between words. The model: “1. Creates a query, key, and value vector for each token in the input sequence. Calculates the similarity between the query vector from step one and the key vector of every other token by taking the dot product of the two vectors. Generates normalized weights by feeding the output of step 2 into a softmax function. Generates a final vector, representing the importance of the token within the sequence by multiplying the weights generated in step 3 by the value vectors of each token.” Ruby, supra note 27. The AI system contextualizes words by using tokens, which are word groupings that help to quantify the context of the specific word type of interest, to simplify the process by transforming arbitrary length vectors into a single fixed length vector. See Smith, supra note 34, at 71–72. The input, just like the “red apple card” now has a value the software can understand.
35
Based on the task the GPT models is trying to accomplish, the desired output format can be specified earlier and produce responses that follow this requirement. Chengwei Qin et al., Is ChatGPT a General-Purpose Natural Language Processing Task Solver, arXiv (Nov. 19, 2023), https://arxiv.org/abs/2302.06476 [hereinafter Is ChatGPT a General-Purpose NLP Task Solver]. The models have a decoder to generate the output that “allows the model to differentially weight parts of the sequence to infer meaning and context.” Ruby, supra note 27. AI models, like ChatGPT, have “comparison data based on human preferences” and reward models that are trained and optimized using reinforcement learning in order to generate “high quality responses to human input, reject[] inappropriate questions, and self-correct[] previous errors based on subsequent conversations.” Is ChatGPT a General-Purpose NLP Task Solver, supra note 35, at 1.
36
See Ruby, supra note 27 (“[T]he encoder leverages masked-language-modeling to understand the relationship between words and produce more comprehensible responses.”). But see Smith, supra note 34, at 74 (“This is not nearly the whole story of NLP; there is much more to be said about approaches to dealing with natural language syntax, semantics, and pragmatics, and how we operationalize tasks of understanding and production that humans perform . . . to design algorithms.”).
37
A natural language processing system starts at the word level to determine its meaning and then transition to the sentence level to hone in on its the grammar, word order, and meaning of the sentence before determining the context of the overall environment. Gobinda G. Chowdhury, Natural Language Processing, 37 Ann. Rev. Info. Sci. & Tech. 51, 55 (2003). The context is then used in a natural language interface when the interface “accepts query statements or commands in natural language and sends data to some system, typically a retrieval system, which then provides appropriate responses to the commands or query statements.” Id. at 66.
38
Ruby, supra note 27.
39
Natural language processing programs are built using a combination of examples of inputs and outputs to a task and generalizing the pairs. See Smith, supra note 34, at 68–69. From these pairs, any information that is discovered about one word can be transferred to similar words. Id.
40
E.g., LegalZoom.com, Inc. v. N.C. State Bar, No. 11 CVS 15111, 2012 WL 3678650, at *2 (N.C. Super. Ct. Aug. 27, 2012); Complaint at ¶¶ 20-26, LegalZoom.com, Inc. v. N.C. State Bar, 2011 WL 8424700 (N.C. Super. Ct. 2011) (No. 11CVS15111) [hereinafter N.C. Complaint].
41
N.C. Complaint, supra note 40, at ¶ 10; Isaac Figueras, The LegalZoom Identity Crisis: Legal Form Provider or Lawyer in Sheep’s Clothing?, 63 Case W. Rsrv. L. Rev. 1419, 1425 (2013).
42
See Figueras, supra note 41, at 1425.
43
Id.
44
N.C. Complaint, supra note 40, at ¶ 10.
45
Compare Figueras, supra note 41, at 1425 (“LegalZoom . . . uses questionnaires created by customers in order to help generate legal documents[,] . . . then prints the documents and mails them to you with instructions regarding how to finalize this document.”), with Hughes, supra note 24 (“Generative AI uses various machine learning techniques, such as GANs, VAEs or LLMs, to generate new content from patterns learned from training data. These outputs can be text, images, music or anything else that can be represented digitally.”); see generally The 7 Stages of the Future of Evolution of Artificial Intelligence, Recode (Feb. 28, 2021), https://recodeminds.com/blog/the-7-stages-of-the-future-evolution-of-artificial-intelligence/ (describing the different complexity levels of AI that accomplishes process automation in Stage 1, and machines with the “capacity to reason, interact and, deal with other machines and humans too” in Stage 4).
46
E.g., Janson v. LegalZoom.com, Inc., 802 F. Supp. 2d 1053 (W.D. Mo. 2011); LegalZoom.com, Inc. v. N.C. State Bar, No. 11 CVS 15111, 2014 WL 1213242 (N.C. Super. Ct. Mar. 24, 2014); Lowry v. LegalZoom.com, Inc., No. 4:11CV02259, 2012 WL 2953109 (N.D. Ohio July 19, 2012); see also In re Peterson, No. 19-24045, 2022 WL 1800949 (Bankr. D. Md. June 1, 2022) (where another program, Upsolve, faced UPL allegations).
47
See Terry Carter, LegalZoom Business Model OK’d by South Carolina Supreme Court, ABA J. (Apr. 25, 2014, 10:20 PM), https://www.abajournal.com/news/article/legalzoom_business_model_okd_by_south_carolina_supreme_court.
48
Medlock v. LegalZoom.com, Inc., Case No. 2012-208067, 2013 S.C. Lexis 362, at *15 (S.C. Oct. 18, 2013).
49
Stephanie Wilkins, ChatGPT Is Impressive, but Can (and Should) it be Used in Legal?, Law.com (Dec. 15, 2022, 1:25 PM), https://www.law.com/legaltechnews/2022/12/15/chatgpt-is-impressive-but-can-and-should-it-be-used-in-legal/.
50
Id.
51
Id.
52
Id. (internal quotation marks omitted).
53
Samantha Murphy Kelly, ChatGPT Passes Exams from Law and Business Schools, CNN Bus., https://www.cnn.com/2023/01/26/tech/chatgpt-passes-exams/index.html (last updated Jan. 26, 2023, 1:35 PM).
54
Daniel Van Boom, ChatGPT Can Pass the Bar Exam. Does That Actually Matter?, CNET (Mar. 19, 2023, 1:55 PM), https://www.cnet.com/tech/chatgpt-can-pass-the-bar-exam-does-that-actually-matter/.
55
Derek A. Denckla, Nonlawyers and the Unauthorized Practice of Law: An Overview of the Legal and Ethical Parameters, 67 Fordham L. Rev. 2581, 2582 (1999).
56
Id.
57
Id.
58
See Model Rules of Pro. Conduct Preamble & Scope cmt. 20 (Am. bar Ass’n 1983) (“The Rules are designed to provide guidance to lawyers and to provide a structure for regulating conduct through disciplinary agencies.”).
59
See, e.g., Eugene R. Gaetke, Why Kentucky Should Adopt the ABA’s Model Rules of Professional Conduct, 74 Ky. L.J. 581, 583–84 (1986).
60
Model Rules of Professional Conduct, Am. Bar Ass’n, https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/ (last visited Apr. 5, 2023).
61
ABA Commission on Ethics 20/20, Am. Bar Ass’n, https://www.americanbar.org/groups/professional_responsibility/committees_commissions/aba-commission-on–ethics-20-20/ (last visited Apr. 5, 2023).
62
See ABA Comm. on Ethics 20/20, Revised 107A (2013) (providing a resolution for foreign lawyers with in-house counsel legal work based on the growth of technology); ABA Comm. on Ethics 20/20, Revised 105A (2011) (providing comments related to the growth of technology and confidentiality).
63
Priorities & Initiatives, Am. Bar Ass’n, https://www.americanbar.org/groups/professional_responsibility/committees_commissions/aba-commission-on–ethics-20-20/priorities_policy/ (last visited Apr. 5, 2023) (displaying the Commission’s priorities and initiatives of issued papers).
64
ABA Commission on Ethics 20/20, supra note 61.
65
See Lucian T. Pera, Grading ABA Leadership on Legal Ethics Leadership: State Adoption of the Revised ABA Model Rules of Professional Conduct, 30 Okla. City U. L. Rev. 637, 640 (2005); Alphabetical List of Jurisdictions Adopting Model Rules, Am. Bar, https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/alpha_list_state_adopting_model_rules/?login (last updated Mar. 28, 2018).
66
See Pera, supra note 65, at 640.
67
Multistate Professional Responsibility Examination, Nat’l Conf. of Bar Examiners, https://www.ncbex.org/exams/mpre/ (last visited Apr. 5, 2023) (Wisconsin and Puerto Rico are the only two jurisdictions that do not require the MPRE for admission to the bar).
68
Another paper of interest could include the confidentiality implications from the use of client information when a machine learning AI program uses proprietary facts to help a lawyer write a brief.
69
See discussion supra notes 40–48.
70
Model Rules of Pro. Conduct r. 5.5 (Am. Bar Ass’n 1983).
71
See Restatement (Third) of the Law Governing Lawyers § 4 (Am. L. Inst. 2000) (“A person not admitted to practice as a lawyer . . . may not engage in the authorized practice of law . . . .”).
72
Model Rules of Pro. Conduct r. 5.5 cmt. 2 (Am. Bar Ass’n 1983).
73
See In re Unauthorized Prac. of Law Rules Proposed by the S.C. Bar, 309 S.C. 304, 305, 422 S.E.2d 123, 124 (1992) (“[I]t is neither practicable nor wise to attempt a comprehensive definition by way of a set of rules.”).
74
See, e.g., S.C. Code Ann. § 40-5-310 (Supp. 2022) (“No person may either practice law or solicit the legal cause of another person or entity in this State unless he is enrolled as a member of the South Carolina Bar . . . or otherwise authorized to perform prescribed legal activities by action of the Supreme Court of South Carolina.”). The punishments for the unauthorized practice of law can be criminal. Md. Code Ann. bus Occ. & Prof., § 10-606 (2008) (UPL violations punishable as a misdemeanor with a maximum fine of $5000 and maximum imprisonment of one year); Alaska Stat. § 34-3-1 (2022) (UPL violations punishable as a misdemeanor with a maximum fine of $500 and maximum imprisonment of six months); R.I Gen. Laws § 11-27-14 (2023) (UPL violations punishable by a maximum fine of $500 or imprisonment of one year with increased penalties for repeat offenders).
75
See Delaware State Bar Ass’n v. Alexander, 386 A.2d 652 (Del. 1978); see also S.C. Code Ann. § 40-5-310 (“The type of conduct that is the subject of any charge . . . must have been defined as the unauthorized practice of law by the Supreme Court of South Carolina prior to any charge being filed.”).
76
See generally Appendix A: State Definitions of the Practice of Law, Am. Bar Ass’n, https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/model-def_migrated/model_def_statutes.authcheckdam.pdf (last visited Apr. 6, 2023).
77
In re Unauthorized Practice of Law, 422 S.E.2d 123, 124 (S.C. 1992).
78
See Ohio State Bar Ass’n v. Home Advoc. Trs., LLC, 152 Ohio St. 3d 60, 63, 92 N.E.3d 862, 865 (2017) (drafting and filing documents in Ohio courts); Ky. State Bar Ass’n v. Cent. Ky. Enters., Inc., 503 S.W.2d 483, 483 (Ky. Ct. App. 1972) (drafting of a deed); Atty. Grievance Comm’n of Md. v. Hallmon, 343 Md. 390, 397, 681 A.2d 510, 514 (1996) (preparation of legal documents). But see State ex rel. Indiana State Bar Ass’n v. Indiana Real Est. Ass’n, 191 N.E.2d 711, 715 (Ind. 1963) (“[F]illing in of blanks in legal instruments, prepared by attorne[y]s, which require only the use of common knowledge regarding the information to be inserted in said blanks, and general knowledge regarding the legal consequences involved, does not constitute the practice of law.”).
79
See In re Reynoso, 477 F.3d 1117, 1125 (9th Cir. 2007) (“the practice of law … includes legal advice and counsel and the preparation of legal instruments and contracts”); Clark v. Gannett Co., Inc., 122 N.E.3d 376, 383, 389 (Ill. App. Ct. 2018) (attorney practiced law in an unauthorized jurisdiction by “calling the shots” and not signing any pleadings); Hous. Auth. of City of Charleston v. Key, 352 S.C. 26, 28, 572 S.E.2d 284, 285 (2002) (Paralegal engaged in the unauthorized practice of law by preparing pleadings and had them signed by the plaintiffs as pro se litigants).
80
See In re Arons, 756 A.2d 867, 868-69 (Del. 2000) (non-attorney with special knowledge engaged in UPL due to representation of families with disabled children at “due process” hearings pursuant to the federal Individuals with Disabilities Education Act); Toledo Bar Ass’n v. Joelson, 114 Ohio St. 3d 425, 426, 872 N.E.2d 1207, 1208 (2007) (“The practice of law is not limited to appearances in court. It also embraces the preparation of papers that are to be filed in court on another’s behalf and that are otherwise incident to a lawsuit.”).
81
See Carter, supra note 47.
82
See Daniel Fisher, LegalZoom Settles Fight with North Carolina Bar Over Online Law, Forbes (Oct. 22, 2015, 2:16 PM), https://www.forbes.com/sites/danielfisher/2015/10/22/legalzoom-settles-fight-with-north-carolina-bar-over-online-law/?sh=5c255d433eb2; see also Debra C. Weiss, LegalZoom Can Continue to Offer Documents in Missouri Under Proposed Settlement, ABA J. (Aug. 23, 2011, 12:32 PM), https://www.abajournal.com/news/article/legalzoom_can_continue_to_offer_documents_in_missouri_under_proposed_settle.
83
How Courts Work, Am. Bar Ass’n (Sept. 9, 2019), https://www.americanbar.org/groups/public_education/resources/law_related_education_network/how_courts_work/court_role/.
84
U.S. Const. pmbl.
85
Id.
86
See Legal Servs. Corp., The Justice Gap: Measuring the Unmet Civil Legal Needs of Low-Income Americans 14 (2017).
87
See id. at 14, 30.
88
See id. at 49–52.
89
Rebecca Buckwalter-Poza, Making Justice Equal, Ctr. for Am. Progress (Dec. 8, 2016), https://www.americanprogress.org/article/making-justice-equal/ (example of Mary Hicks); see Legal Servs. Corp., supra note 86, at 23 (29% of households experience rental housing issue in 2017).
90
Buckwalter-Poza, supra note 89.
91
Id.
92
Id.
93
Legal Servs. Corp., supra note 86, at 14.
94
Id. at 29.
95
Nick Rishwain, How Courts Can Increase Access to Justice by Adopting Better Technology, GPSolo, Jan.–Feb. 2019, at 40.
96
Id. at 41–42 (detailing different technologies that are utilized to aid in lowering costs as an obstacle to access to justice, such as CourtCall, e-filing, and Experts.com).
97
See id. at 42.
98
Legal Servs. Corp., supra note 86, at 13.
99
Id. at 33.
100
See Artificial Intelligence: What It Is and Why It Matters, supra note 32.
101
See id.
102
For example, a user could input “what is a breach of duty” into an AI model to get clarification on a legal issue presented in an AI response. See generally ChatGPT, OpenAI, https://chat.openai.com/ (last visited Feb. 4, 2024). Different AI models are used to help with learning new information by providing definitions of key words and personalized feedback and assistance. See W. Ian O’Byrne, 4 Ways That AI Can Help Students, The Conversation (Apr. 11, 2023, 8:06 AM), https://theconversation.com/4-ways-that-ai-can-help-students-200973.
103
See Buckwalter-Poza, supra note 89; ChatGPT, OpenAI, https://chat.openai.com/ (last visited Jan. 8, 2024).
104
ChatGPT, supra note 103.
105
Id.
106
See id.
107
Jona Goldschmidt, The Pro Se Litigant’s Struggle for Access to Justice: Meeting the Challenge of Bench and Bar Resistance, 40 Fam. Ct. Rev. 36, 37 (2002).
108
See id. at 38.
109
Id.
110
About SRLN – 2020 Report, Self-Represented Litig. Network (Jan. 12, 2024), https://www.srln.org/node/21/about-srln.
111
Id.
112
See Pub. Serv. Comm’n of S.C., So, I’m Going to Represent Myself Before the PSC. What Next?: A Pro Se Litigant Guide, App. B (2021), https://psc.sc.gov/sites/psc/files/Documents/Pro%20Se%20Litigant/FINAL%20Amended%20Pro%20Se%20Litigant%20Guide.pdf.
113
See U.S. Ct. of Fed. Claims, A Guide for Self-Representation (2022), https://www.uscfc.uscourts.gov/sites/default/files/pro_se_guidebook_20221118.pdf; U.S. Dist. Ct. Dist. of S.C., Information on Representing Yourself in a Civil Action (Non-Prisoner) (2021), http://www.scd.uscourts.gov/DOCS/PROSE.pdf.
114
See Common Legal Topics, S.C. Bar Ass’n, https://www.scbar.org/public/get-legal-help/common-legal-topics/(last visited Apr. 9, 2023).
115
Goldschmidt, supra note 107, at 47–48.
116
ABA Survey Finds 1.3M Lawyers in the U.S., Am. Bar Ass’n (June 20, 2022), https://www.americanbar.org/news/abanews/aba-news-archives/2022/06/aba-lawyers-survey/.
117
See id.; Legal Servs. Corp., supra note 86, at 44.
118
Legal Servs. Corp., supra note 86, at 14. Similarly, the middle class does not receive much help as 40 to 60 percent of their legal needs also go unmet. Jennifer S. Bard & Larry Cunningham, Opinion: The Legal Profession is Failing Low-Income and Middle-Class People. Let’s Fix That, The Wash. Post (June 5, 2017, 9:52 AM), https://www.washingtonpost.com/opinions/the-legal-profession-is-failing-low-income-and-middle-class-people-lets-fix-that/2017/06/02/e266200a-246b-11e7-bb9d-8cd6118e1409_story.html.
119
See What Is a Lawyer?, U.S. News, https://money.usnews.com/careers/best-jobs/lawyer (last visited Feb. 4, 2024).
120
Catherine Albiston et al., Making Public Interest Lawyers in a Time of Crisis: An Evidence-Based Approach, 34 Geo. J. Legal Ethics 223, 233 (2021).
121
Id.
122
Id. at 236; see also Juliet R. Bailin, At HLS, A Tough Path to Public Interest, The Harv. Crimson (Oct. 29, 2012), https://www.thecrimson.com/article/2012/10/29/hls-public-service/ (“One of the ironies in life . . . is you work four times as hard to get a quarter of the money in public interest.”).
123
David C. Vladeck, Hard Choices: Thoughts for New Lawyers, 10 Kan. J.L. & Pub. Pol’y 351, 352 (2000).
124
Id. (“Even assuming that the lower figure of the number of lawyers in the United States, 752,000, is correct . . . the fact remains that fewer than one percent [sic] of our nation’s lawyers provide legal services to the poor.”).
125
Kristen Sonday, Tech-Enabled A2J: How Tech is Helping Pro Se Litigants Navigate the Courts, Thomson Reuters (Aug. 17, 2020), https://www.thomsonreuters.com/en-us/posts/legal/tech-enabled-a2j-pro-se-litigants/.
126
Id.
127
Id.
128
See Jack Knight, The Bases of Cooperation: Social Norms and the Rule of Law, 154 J. Inst. & Theoretical Econ. 754, 762 (1998).
129
Amy J. Schmitz, Measuring “Access to Justice” in the Rush to Digitize, 88 Fordham L. Rev. 2381, 2396 (2020) (“Negative consequences emerge when individuals no longer trust the role of the law or communal institutions charged with protecting justice.”).
130
See Knight, supra note 128, at 762.
131
Michael P. Stone & Thomas J. Miceli, The Impact of Frivolous Lawsuits on Deterrence: Do They Have Some Redeeming Value?, 10 J.L. Econ. & Pol’y 301, 302 (2014).
132
See Stephen J. Ware, Is Adjudication a Public Good? “Overcrowded Courts” and the Private Sector Alternative of Arbitration, 14 Cardozo J. Conflict Resol. 899, 899 (2013).
133
See id.
134
See Jay M. Feinman, Five Myths About Lawsuits, The Wash. Post (July 23, 2020, 10:05 AM), https://www.washingtonpost.com/outlook/five-myths/five-myths-about-lawsuits/2020/07/23/8006d532-c169-11ea-b4f6-cb39cd8940fb_story.html.
135
Model Rules of Pro. Conduct r. 3.1 (Am. Bar. Ass’n 1983) (placing a duty on lawyers to not bring frivolous issues that are not “good faith argument for an extension, modification or reversal of existing law.”); see Wesley A. Cann Jr., Frivolous Lawsuits – The Lawyer’s Duty to Say “No,” 52 U. Colo. L. Rev. 367, 367 (1981).
136
See Fed. R. Civ. P. 11(b)(2).
137
Michael D. Roundy, The Proper Approach to Pro Se Litigants, Am. Bar Ass’n (July 30, 2020), https://www.americanbar.org/groups/litigation/committees/pretrial-practice-discovery/practice/2020/proper-approach-to-pro-se-litigants/.
138
See Feinman, supra note 134 (where arbitration clauses and legislative immunities cause a myth that individuals can always sue).
139
See Cann Jr., supra note 135, at 368; Ware, supra note 132, at 904.
140
Cann Jr., supra note 135, at 368–69.
141
For example, automobiles replacing horses as the main mode of transportation was believed to be good for pollution, “and now automobiles are one of the leading cause[s] of the planet’s Co2 pollution and other serious problems.” Peter Milsom, Pollution – Why We Replaced Horses with Automobiles, Green Project Mgmt. (May 13, 2019), https://blog.greenprojectmanagement.org/index.php/2019/05/13/pollution-why-we-replaced-horses-with-automobiles/.
142
See generally id.; Liraz Margalit, Chatbots: The Illusion of Companionship Without the Demands of Friendship, CMSWire (Oct. 22, 2022), https://www.cmswire.com/customer-experience/chatbots-the-illusion-of-companionship-without-the-demands-of-friendship/.
143
Molly Bohannon, Lawyer Used ChatGPT in Court—And Cited Fake Cases. A Judge Is Considering Sanctions, Forbes (June 8, 2023, 2:06 PM), https://www.forbes.com/sites/mollybohannon/2023/06/08/lawyer-used-chatgpt-in-court-and-cited-fake-cases-a-judge-is-considering-sanctions/?sh=4de2c7167c7f.
144
See In re Unauthorized Prac. of Law Rules Proposed by the S.C. Bar, 309 S.C. 304, 305, 422 S.E.2d 123, 124 (1992).
145
LegalForce RAPC Worldwide, P.C. v. DeMassa, 532 F. Supp. 3d 856, 861 (N.D. Cal. 2021) (citing Baron v. City of Los Angeles, 2 Cal.3d 535, 542, 469 P.2d 353, 357 (Cal. 1970).
146
Id.
147
Id. (quoting Gardner v. Conway, 234 Minn. 468, 481, 48 N.W.3d 788, 796 (1951)) (internal quotation marks omitted). One court has gone so far as to say “[t]he preparation for clients documents requiring familiarity with legal principals beyond the ken of the ordinary layman constitutes the unauthorized practice of law under Pennsylvania law.” In re Dunkle, 272 B.R. 450, 453 (Bankr. W.D. Pa. 2002) (quoting In re Maloney, 249 B.R. 71, 76 (M.D. Pa. 2002)) (internal quotation marks omitted).
148
Disciplinary Couns. v. Ward, 155 Ohio St. 3d 488, 490, 122 N.E.3d 168, 171 (2018) (quoting Lorain Cty. Bar Ass’n v. Kocak, 121 Ohio St. 3d 396, 399, 904 N.E.2d 885, 889 (2009)).
149
Lucas Subway MidMo, Inc. v. Mandatory Poster Agency, Inc., 524 S.W.3d 116, 124–25 (W.D. Mo. Ct. App. 2017).
150
Id. at 123 (quoting Mo. Rev. Stat. Ann. § 484.010.1 (2023)).
151
Rule 11 concerns an attorney signing “[e]very pleading, written motion, and other paper” presented to the court and preparing the papers submitted to the court for proper, nonfrivolous purposes. See Fed. R. Civ. P. 11(a)–(b).
152
FIA Card Servs., N.A. v. Pichette, 116 A.3d 770, 780 (R.I. 2015).
153
Id. at 781.
154
Id. at 784.
155
See supra notes 105–07.
156
ChatGPT, supra note 103.
157
Id.
158
See David Goguen, Finding Your Way Through Court: Steps in a Lawsuit, Lawyers, https://legal-info.lawyers.com/research/steps-in-a-lawsuit.html (last updated July 23, 2020); Mo. Rev. Stat. Ann. § 484.010.1 (2023).
159
See Drafting Complaints: Start off on the Right Foot, Sheppard Mullin, https://www.sheppardmullin.com/media/article/1633_Drafting%20Complaints.pdf (last visited Jan. 15, 2024).
160
See, e.g., State Bar v. Guardian Abstract & Title Co., Inc., 91 N.M. 434, 440, 575 P.2d 943, 949 (1978).
161
532 F. Supp. 3d 856 (N.D. Ca. 2021).
162
Id. at 865–66.
163
Id. at 866.
164
Id.
165
Id.
166
Id.
167
See id.
168
See id.
169
ChatGPT, supra note 103.
170
Id.
171
Barlow F. Christensen, The Unauthorized Practice of Law: Do Good Fences Really Make Good Neighbors – Or Even Good Sense?, 5 Am. Bar. Found. Rsch. J. 159, 164 (1980).
172
See Villasenor, supra note 13 (“There will be people who exploit the nearly frictionless ability to automatically generate legal complaints to rapidly flood court systems in multiple jurisdictions with frivolous AI-written lawsuits.”).
173
Amy B. Cyphert, A Human Being Wrote This Law Review Article: GPT-3 and the Practice of Law, 55 U.C. Davis L. Rev. 401, 437 (2021) (“If a tool like GPT-3 is to be used successfully in addressing the access to justice gap, it must be reevaluated and updated with an eye toward more culturally competent design.” . . . “[L]awyers will not be able to remove bias from AI systems, no matter how technically competent they become or how rigorously they supervise the systems.”); see also Molly Bohannon, Lawyer Used ChatCPT in Court – And Cited Fake Cases. A Judge Is Considering Sanctions, Forbes (Jun. 8, 2023, 2:06 PM), https://www.forbes.com/sites/mollybohannon/2023/06/08/lawyer-used-chatgpt-in-court-and-cited-fake-cases-a-judge-is-considering-sanctions/?sh=8769c077c7f3 (showing the potential of lawyers compromising their competency by using ChatGPT).
174
See Lance Elliot, ChatGPT and Other AI Programs Aid and Muddle Access to Justice as Non-Lawyers Seek Their Advice, Jurist (Mar. 7, 2023, 12:35 PM), https://www.jurist.org/commentary/2023/03/eliot-ai-future-lawyering/ (“[T]he future will almost inevitably cause a change [to UPL provisions] to occur since it is highly likely that generative AI will improve and eventually embody the legal field in a robust and lawyering capacity.”).
175
Mahoning Cty. Bar Ass’n v. Amatore, 164 N.E.3d 458, 459 (Ohio 2021).
176
State ex rel. Fla. Bar v. Sperry, 140 So. 2d 587, 595 (Fla. 1962), vacated sub nom. Sperry v. State of Fla. ex rel. Fla. Bar, 373 U.S. 379 (1963).
177
Erika C. Birg, Lawyers on the Road: The Unauthorized Practice of Law and the 2004 Presidential Election, 9 Tex. R. L. & Pol. 305, 319 (2005).
178
Model Rules of Pro. Conduct r. 1.1 (Am. Bar Ass’n 1983).
179
Id. at r. 1.6.
180
Id. at r. 1.5.
181
See Birg, supra note 177, at 310–12.
182
George C. Leef, Lawyer Fees to High? The Case for Repealing Unauthorized Practice of Law Statutes, Regul., Winter 1997, at 33.
183
See id.; Susan D. Hoppock, Enforcing Unauthorized Practice of Law Prohibitions: The Emergence of the Private Cause of Action and Its Impact on Effective Enforcement, 20 Geo. J. Legal Ethics 719, 726 (2007).
184
See Leef, supra note 182, at 33, 38.
185
See discussion supra notes 3–148.
186
See Vladeck, supra note 123, at 353 (“Unless a lawyer is willing to take [a $20,000] case on for a contingency fee – which is hardly a certainty given the complexity of the case and the modest size of the potential recovery – they might well be out of luck in finding a lawyer willing to help them.”).
187
See id. at 356 (“But more often than not, restrictions on non-lawyer practice end up hurting only the public without even protecting the profession’s pocketbook.”).
188
See Leef, supra note 177, at 33, 38.
189
Id. at 184.
190
Margalit, supra note 142 (“Communication with a bot is different . . . [y]ou can achieve your goal (getting help, information, even a feeling of companionship) with no immediate ‘cost.’”); see Lennart Seitz et al., Can We Trust a Chatbot Like a Physician? A Qualitative Study on Understanding the Emergence of Trust Toward Diagnostic Chatbots, 165 Int’l J. Hum. Comput. Stud. 102848, 1 (2022) (noting that healthcare chatbots “offer a simple and efficient form of information, empowering patients to engage in decision-making processes and self-care.”).
191
See John Leubsdorf, Legal Malpractice and Professional Responsibility, 48 Rutgers L. Rev. 101, 102–03 (1995).
192
See Tambiama Madiega, European Parliamentary Rsch. Serv. Artificial Intelligence Liability Directive 4 (2023), https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/739342/EPRS_BRI(2023)739342_EN.pdf (“[T]here is no comprehensive federal legislation on AI and addressing liabilities in the US to date.”); see also Robayet Syed, So Sue Me: Who Should Be Held Liable When AI Makes Mistakes?, Monash Univ. (Mar. 29, 2023), https://lens.monash.edu/@politics-society/2023/03/29/1385545/so-sue-me-wholl-be-held-liable-when-ai-makes-mistakes (“But AI systems are considered property, and don’t have the same legal rights and responsibilities as humans or legal entities.”).
193
Leubsdorf, supra note 191, at 102–03.
194
Id. at 108.
195
Id. at 109.
196
Id. at 112.
197
Hoppock, supra note 183, at 730.
198
Id.
199
See id. at 732–33.
200
Id.
201
David B. Wilkins, Who Should Regulate Lawyers?, 105 Harv. L. Rev. 799, 812 (1992).
202
Id.
203
Hoppock, supra note 183, at 733.
204
See Wilkins, supra note 201, at 835.
205
See id.
206
See Leubsdorf, supra note 191, at 102.
207
See Corinne Cath et al., Artificial Intelligence and the ‘Good Society’: The US, EU, and UK Approach, 24 Sci. Eng’r Ethics 505, 516 (2018).
208
See id.
209
See Browning & Lecun, supra note 5.
210
Id.
211
Id.
212
Id.
213
Eirini Ntoutsi et al., Bias in Data-Driven Artificial Intelligence Systems—An Introductory Survey, 10 WIREs Data Mining & Knowledge Discovery 1356, 2 (2020).
214
Id. at 3.
215
Id.
216
Id.
217
See Debra Lyn Bassett, Deconstruct and Superstruct: Examining Bias Across the Legal System, 46 U.C. Davis L. Rev. 1563, 1564 (2013).
218
Id.
219
See Model Rules of Pro. Conduct r. 1.7 (Am. Bar Ass’n 1983).
220
Id. at cmt. 2.
221
Id. at r. 5.5(a).
222
Id. at cmt. 2.
223
Suggestion from author after analyzing wording in a multitude of comments from the Model Rules of Professional Conduct. See generally Model Rules of Pro. Conduct (Am. Bar Ass’n 1983).
224
See Wilkins, supra note 201, at 835.
225
See discussion supra notes 144–154.
226
See Legal Servs. Corp., supra note 86, at 14.
227
Marcy Strauss, Toward a Revised Model of Attorney-Client Relationship: The Argument for Autonomy, 65 N.C. L. Rev. 315, 338 (1987).
228
See id.
229
See id.
230
Leubsdorf, supra note 191, at 102.
231
Fed. R. Civ. P. 11.
232
See Model Rules of Pro. Conduct r. 5 cmt. 2 (Am. Bar Ass’n 1983).
233
Id. (“This Rule does not prohibit a lawyer from employing the services of paraprofessionals . . . , so long as the lawyer supervises the delegated work and retains responsibility for the work.”).