top of page
Writer's pictureCynthia Sharp

ChatGPT: What Lawyers Need to Know Before Using AI

By Rebecca Howlett and Cynthia Sharp


Change is not merely necessary to life—it is life. —Futurist Alvin Toffler


What Is ChatGPT and Why Should Lawyers Care?


By the time you read this, this article will already be out-of-date. Rapid advancements in generative artificial intelligence (AI), such as ChatGPT, are happening minute by minute. Indeed, the emergence of ChatGPT has made an immediate global impact. Technology experts are comparing this to the dawn of the World Wide Web or the advent of social media. In its first two months of launch, ChatGPT generated more than 100 million users—setting the record for the fastest-growing consumer application ever.


The co-authors of this article are personally convinced that this type of technology is going to drastically and fundamentally alter how we approach the practice of law and our lives in general. For example, the World Economic Forum estimates that AI will eliminate 85 million jobs by 2025; however, it is also estimated to create 97 million new jobs at the same time. Overall, evolving alongside the emergence and integration of AI within the legal field will be key if attorneys want to stay relevant—and even ethically compliant.


We recently interviewed fellow attorney and ChatGPT expert Marc E. Hoag on our podcast, The Legal Mindset Corner, about the implications of AI on the practice of law. As a San Francisco–based solo AI attorney, Marc is on the front lines of helping attorneys effectively navigate and leverage artificial intelligence in the legal field. Like your co-authors, Hoag is convinced that AI has the potential to completely change the game for lawyers—and humanity.


When asked why lawyers should care about AI such as ChatGPT, Hoag responded, “Any attorney or law practice, any company, any business, any profession that does not embrace AI in some capacity will be dead or dying in the next two to three years.” That’s a bold prediction, but one your co-authors wholeheartedly agree with. To stay up-to-date on the latest AI developments, follow Future Perfect, which offers daily highlights of the latest AI news and tools.


Given the present state of play, we expect that AI technology such as ChatGPT will be a mainstay of our law practices and lives in just a couple of years. With that said, we want to empower lawyers to maximize the positive benefits of employing ChatGPT, which necessarily includes understanding the ethical implications—of which there are many!


Personal Uses of Generative AI

Obviously, the thrust of this article is how attorneys can leverage ChatGPT in their legal practices, but the uses of generative AI go far beyond the professional context. The co-authors use it all the time in our personal lives as well! For example, Cindy recently decided to make some positive lifestyle changes to her diet, given a family history of high cholesterol. Naturally, she queried ChatGPT to create a heart-healthy menu for a 70-year-old who likes dessert.


We have also used ChatGPT to suggest book recommendations, write jokes and satire pieces for family celebrations, as well as draft personal correspondence, including difficult-to-navigate conversations such as ending a friendship that no longer serves our highest and best self.


The world is your oyster! And if you don’t like or are allergic to shellfish, you can use ChatGPT to suggest a more palatable metaphor: The world is your stage! The sky’s the limit! You get the idea.


Leveraging ChatGPT in the Legal Field


We know of one lawyer who uses ChatGPT so frequently that her office refers to it as “Gloria.” After Gloria completes the first draft of an “assignment,” the lawyer provides the finishing touches, a process that is significantly less time-consuming than facing the tyranny of the blank page. Examples of projects assigned to Gloria include drafting blogs, LinkedIn posts, and FAQs for the website. The attorney related that utilizing this resource for routine tasks has resulted in an increase in personal productivity and that the attorney now has more time to explore higher-level projects that require creativity, wisdom, and judgment.


An excellent resource for lawyers exploring the use of ChatGPT is OpenAI’s recently released guide to GPT Best Practices, which provides basic information as well as detailed suggestions on achieving better results. Those beginning their AI journey also need to understand that there is an art to framing requests that will get the most helpful responses from ChatGPT and other chatbots. This evolving skill set is being referred to as “prompt engineering.” Articles that we recommend for more information on this topic include David Nielo’s “11 Tips to Take Your ChatGPT Prompts to the Next Level” (Wired, Mar. 26, 2023) and Kristen Barkved’s “6 ChatGPT Prompts for Lawyers” (Clio, June 15, 2023).


As discussed in later sections of this article, attorneys using ChatGPT to draft legal documents must be extra vigilant in verifying the accuracy of the output. Think of anything generated by ChatGPT as being produced by an advanced law student who requires close supervision.


That being said, a number of AI platforms have been developed specifically for lawyers and may produce accurate legal-related output. Although we have not independently reviewed the platforms, we understand that their capabilities include contract analysis and review, client communications, due diligence and eDiscovery, drafting and editing legal documents, document management, and legal research. Because the field is evolving at warp speed, we hesitate to mention any specific products (at least for now) and suggest that our readers conduct their own independent research.


Possible Pitfalls of ChatGPT in the Practice of Law


Danger, Will Robinson! — Robot, Lost in Space


Before you start fully integrating ChatGPT into your law practice, a word of warning. Diligently cultivate full awareness of the implications of this technology and its inherent risks, and prioritize ethical use from the outset!


Although the possibilities are endless for the positive benefits of AI in our law practices and lives, the possibilities are also endless for what could go awry. This type of technology is extremely novel, and there are many unknowns. In effect, we are active participants in a global experiment. As Hoag suggests, “It’s sort of like we’re allowing public, above-ground nuclear testing. It’s like flying a passenger plane full of passengers when it’s still doing test flights.”


Lawyers using generative AI such as ChatGPT for substantive legal work must be mindful of the inherent risk of bias and misinformation. For example, ChatGPT may favor old yet most frequently cited information. This is particularly problematic for attorneys, as these AI tools may overlook new or changing legal interpretations and legal principles.


Be cognizant that any biases or limitations in the data that ChatGPT relies on will be reflected back in its responses, leading to potentially inaccurate answers. Known as a “black box,” ChatGPT can’t necessarily explain how it reached a particular conclusion. Also, understand that small changes in your phrasing of the prompt may yield very different responses.


Further, be aware that ChatGPT has been shown to “hallucinate” and fabricate information when it doesn’t have sufficient data to answer a prompt. In particular, earlier versions of ChatGPT have simply made up names and dates, historical events, and legal case outcomes that never happened or citations that don’t exist!


While that may not inspire confidence, realize that ChatGPT is getting more accurate and advanced by the second. For perspective, ChatGPT-3.5 scored in the tenth percentile on the Uniform Bar Exam, whereas ChatGPT-4 scored in the 90th percentile! Indeed, as AI’s accuracy continues to rapidly and significantly increase, experts warn we may become too trusting of its output.


Going forward, we strongly encourage attorneys to prioritize mindful use of this technology and invest in targeted education and training, including on prompt engineering. Overall, if you choose to use ChatGPT for a substantive legal issue, proceed with caution and ensure you are independently verifying the information.


Potential Ethical Issues


When we delivered our first CLE on the topic of ChatGPT on May 11, 2023, we reported that we knew of no cases relating to attorney misuse of ChatGPT. In the meantime, attorney Steven A. Schwartz has now unfortunately paved the way. He recently filed a brief in federal district court—completely generated by ChatGPT—replete with made-up citations to fake legal opinions. It turns out that neither he nor anyone working for him bothered to double-check the citations, let alone read the so-called cases.


Judge P. Kevin Castel held a hearing on June 8 to determine what sanctions, if any, would be appropriate. To add insult to injury, the name of Schwartz’s partner Peter LoDuca was also on the brief, so he has exposure as well.


In his defense, Schwartz stated, “I heard about this new site, which I falsely assumed was, like, a super search engine.” He admitted to the court that he had neither read the cited cases nor done anything to verify that they were actual cases.


Model Rule 1.1: Competence


Not only is Schwartz subject to sanctions, but he may also be responsible for ethical misconduct. For example, American Bar Association Model Rule of Professional Conduct 1.1 requires attorneys to provide legal services to clients with the requisite knowledge, skill, thoroughness, and preparation necessary for the representation. Schwartz fell short of this standard by relying completely on ChatGPT to produce the finished product. Furthermore, under Comment [8], lawyers have a duty to understand the “benefits and risks associated with relevant technology” if used in their law practice. By his own admission, Schwartz did not fulfill that duty.


“I just never could imagine that ChatGPT would fabricate cases,” Schwartz said. “My assumption was I was using a search engine that was using sources I don’t have access to.” Schwartz swore off using the app in the future and also said he’d taken an AI training course to improve his knowledge of technology.


Another lawyer for the firm, Ronald Minkoff, told the judge that lawyers are notoriously bad with technology. We vehemently disagree with this statement and implore our readers to continue with your quest to master relevant technological advances. Indeed, we encourage lawyers to mindfully explore these AI tools and recommend targeted training before employing AI for substantive legal issues.


Model Rule 1.3: Diligence


Another possible source of violation involves Model Rule 1.3, which requires attorneys to act “with reasonable diligence and promptness in representing a client.” Certainly, ChatGPT could help attorneys satisfy the promptness requirement because, if used correctly, it is a huge time saver. However, relying solely on ChatGPT for legal advice or research without conducting thorough independent analysis is most likely a violation of this rule. It is our obligation to carefully cross-reference the AI-generated output with relevant statutory, regulatory, and case law and to exercise professional judgment every step of the way.


Other Potential Ethical Issues


While it doesn’t appear that the following rules come into play with respect to the Schwartz matter, they are worth considering in this context.


ChatGPT’s web page “What Is ChatGPT?” reveals that they do review conversations and warns, “Please don’t share any sensitive information in your conversations.” Accordingly, attorneys who disclose confidential client information in the course of using ChatGPT have violated Model Rule 1.6(c) because they have failed to take reasonable steps to “prevent the inadvertent or unauthorized disclosure” of such information.


Should law firm retainer agreements be revised to reflect the use of ChatGPT in creating legal documents? Another issue to consider is Model Rule 1.4, which requires transparency with our clients. As an example, the retainer agreement Cindy used in her practice stated that other attorneys and staff may work on the client’s matter. Perhaps our retainer agreements will need to be revised in this context as well. Let us know what you think.


Also worth mentioning are Model Rules 5.1 and 5.3, which establish an attorney’s duty to make “reasonable efforts to ensure” that the conduct of lawyers and nonlawyers over which they have supervisory authority “is compatible with the professional obligations of the lawyer.” This means that legal entities must provide training and also draft an organization-wide policy regarding the use of ChatGPT. Why not let ChatGPT itself provide you with the first draft of that policy? We have even heard of firms that have banned the use of ChatGPT (for now). If you are employed in the legal setting, be sure to familiarize yourself with any AI policy put in force by leadership.


Additional Considerations


Because many firms have already created rules regarding attorney and staff use of ChatGPT, it is essential that employees of law firms and other entities familiarize themselves with the details of any such policy. Furthermore, at least one legal malpractice insurer has distributed a memo to its clients about the potential dangers of using ChatGPT in the law firm setting.

Certainly, many other issues beyond the scope of this article remain to be explored. For the reader’s convenience, we are providing links to other sources that have examined the following topics: unauthorized practice of law, plagiarism and copyright infringement, and defamation.


By the way, if you want to check out whether a document or message sent to you was drafted by ChatGPT, there are several online tools available. The two that we tested are AI Text Classifier (designed by OpenAI) and GPT-4, ChatGPT & AI Detector by ZeroGPT: Detect OpenAI Text. While these and other similar tools may not be foolproof, the results were amazingly accurate when we put them to the test with both human and AI-generated text.


The Perils of Ignorance


Even if you decide to ignore ChatGPT, keep in mind that many of your clients are already aware of and using it. The article How to Use ChatGPT for Legal Advice” (Jim Rogan, GripRoom, Feb. 16, 2013) outlines sample legal questions that can be asked and offers guidance on how to write better prompts. An enterprising student recently used ChatGPT to beat a parking ticket by simply outlining details and requesting it to prepare a letter appealing the fine. Certainly, many consumers of the DIY ilk will consult with ChatGPT with respect to a variety of matters rather than seeking the advice of human counsel.


Moving Forward


While ChatGPT and other AI tools are certainly game changers, human lawyers will continue to be indispensable. The essence of the legal profession encompasses human judgment, ethics, and advocacy, which cannot yet be automated. The role of attorneys will simply continue to evolve as it always has.


The law firm of the future must acknowledge (sooner rather than later) the infinite potential of ChatGPT. By recognizing its capabilities, developing expertise in its applications, and remaining mindful of ethical considerations, the firm will be poised to be a leader in innovation, efficiency, and client service.


Our next column will return the world of the “Legal Burnout Solution” and will discuss how pets can positively benefit our mental health and well-being.


Rebecca Howlett, Esq. and Cynthia Sharp, Esq. are co-founders of The Legal Burnout Solution (legalburnout.com) a community dedicated to the well-being of lawyers. Check out The Legal Mindset Corner, a podcast dedicated to tackling the unique challenges of the legal profession.


Originally published in ABA GPSolo eReport, June 2023 Issue (Vol. 12, No. 11) by the American Bar Association. Reproduced with permission. All rights reserved. This information or any portion thereof may not be copied or disseminated in any form or by any means or stored in an electronic database or retrieval system without the express written consent of the American Bar Association or the copyright holder.


Kommentit


bottom of page