AI in the Legal Profession
Reinhold Leitner/Shutterstock
Artificial intelligence (AI) has become increasingly popular in recent years. The advent of programs such as ChatGPT and Microsoft’s CoPilot has exposed the public to computers doing what was previously thought impossible. Even though legislation has struggled to keep pace with AI and technology, it seems that the legal profession, composed of attorneys and paralegals, has taken to it. AI’s promise to increase efficiency through automating menial tasks is alluring to those in an industry requiring careful consideration of every possibility, document, and word. With that responsibility, many are concerned that the faults of AI could pose serious issues and barriers to the work of attorneys. Lawyers are held to rigid rules by the American Bar Association (ABA), and mistakes could result in the suspension of legal licenses. Questions on whether that risk is worth it will rage on for decades, but to lay the foundation for that debate we must look at how the legal profession currently uses AI.
To begin, it is essential to define AI. There are many preconceptions, fields, and types of AI, but at its core, AI is “a field, which combines computer science and robust datasets, to enable problem-solving” (IBM 2023a). Traditionally, programmers give computers methods or “demands” to solve problems, but the study of AI is the study of making computers “smart” enough to figure out the method themselves. What has been swarming headlines the past few years is a subset of AI called machine learning (ML), which is “the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy” (IBM 2023b). A key factor of ML is the ability of the program to learn and improve itself. Rather than having someone fix the code to make it more efficient, the program can do so by itself. The most obvious example of ML is ChatGPT, the “AI Chatbot.” ChatGPT has been learning to communicate in natural language (the way people speak) by being fed large amounts of data. In all its various forms, ML programs have become the most apparent type of AI program thanks to years of innovation. Since current commercially-available AI don’t have general intelligence like humans, they are considered “narrow AI,” meaning AI that performs specific tasks such as chatting or recommending.
When considering AI use in the legal field, this “narrow AI” is precisely what is used. It has become increasingly common for attorneys and paralegals to use AI to assist them. Specifically, it’s been noted that there are three law-specific categories that AI is being used in:
Document Review
Legal Research
Contract and Legal Document Analysis (Marwaha 2017)
These three categories fundamentally deal with document analysis, an area in which ML has enabled huge improvements in efficiency. In this context, the document refers to actual documents such as tax forms or contracts and text messages, emails, or graphs. Regarding computers, traditional, non-ML programs typically deal with structured data such as stock market prices. This data is already organized for analysis, so the program extracts and analyzes it. Traditional programs have historically had trouble with unstructured data, one example of such being data written in natural languages such as this article. Since the data in an unstructured document isn’t organized, the computer must extract the data from the written word (Bettenbuk 2023).
ML has enabled unstructured data analysis with increasing accuracy. The science behind it is incredibly complex, but on the broadest level, ML programs will analyze a document's textual, visual, and auditory aspects. As it analyzes more documents and gets feedback from another program or a human, it alters how it extracts the information to become increasingly accurate (LinkedIn Staff n.d.). This process is how ML programs and their creators approach document analysis. Document analysis is performed for all three categories listed above. However, slight differences exist in how that data is used once collected.
First, document review is the process of reading documents to find relevant information for a case. Traditionally, paralegals and associates would sift through documents individually, but ML has changed that process. Companies such as Everlaw are developing programs that use ML to assist attorneys in a process called technology-assisted review (TAR). These programs essentially use input, typically a case brief, and sort documents using critical elements from the brief. Oftentimes, the program will also summarize documents and thematically sort them (into groups such as healthcare or safety) to better assist attorneys (Everlaw 2022).
Second, legal research is “the process you use to identify and find the laws — including statutes, regulations, and court opinions — that apply to the facts of your case” (Reuters 2023). Again, legal research previously required associates and paralegals to sift through dense legal documents. That meant looking through books of laws and cases, but digital databases allowed them to use search functions like JSTOR. The problem is that law is an intricate web: often, laws are repealed or amended through other legislative pieces—in other words, they can be very difficult to track. Programs such as Bloomberg’s Brief Analyzer simplify that process. The analyzer reviews case briefs, finds relevant law, then explains why those laws were chosen. In theory, this reduces the time professionals spend doing legal research and allows them to spend more time using that research to adequately build a case (Ambrogi 2020).
Last, contract and legal analysis is the process of reviewing a document, usually a contract, before having a client sign it. Contracts can be huge amounts of work, taking hours to read through a single time. ML programs can review and summarize a document, even stating key problem areas. These programs can also use generative AI, like ChatGPT, to create contracts from scratch based on attorneys’ input. Both these applications can save attorneys massive amounts of time (Silverman 2019).
These three ML applications to law are promising, but of course, they create issues in their own right. ML has many problems at this stage, but the most concerning one revolves around the mistakes AI can make. Lawyers have a massive burden placed upon them: they must be diligent. Failure to review every piece of their case for false information can lead to a loss of profits and potentially a loss of an attorney’s license to practice. While ML’s application to the three categories above could increase efficiency, it requires a large amount of importance to be placed on the work they produce. To that end, it’s possible that an overreliance on ML could lead to disastrous consequences. It’s been well recorded that even well-developed AI such as ChatGPT tends to create hallucinatory citations, misrepresent data, and be factually incorrect about issues (Welborn 2023). There have been a few documented cases of lawyers getting into hot water because of these flaws. The most popular is Mata v. Avianca (2023), in which two New York attorneys cited fake case law given to them by ChatGPT. That incident led to them paying penalty fees of $5,000 (Ryan, Garrett, & Sears 2023). While there have been few other cases as severe as this one, the potential for damage rises as AI continues to be used more in the courtroom.
Though AI can be a helpful tool, it should be used very carefully in sensitive professions such as law. AI’s ability to automate time-intensive tasks such as document review can help lawyers become more efficient than ever, but at the same time, attorneys should be cautious about overreliance on this still relatively new technology.
References
Ambrogi, Bob. 2020. “Bloomberg Law Launches Brief Analyzer, Tool That Uses AI to Review Briefs.” LawSites. February 11, 2020. https://www.lawnext.com/2020/02/bloomberg-law-launches-brief-analyzer-tool-that-uses-ai-to-review-briefs.html.
Bettenbuk, Zoltan. 2023. “Structured Data vs Unstructured Data in Web Scraping.” ScraperAPI. May 23, 2023. https://www.scraperapi.com/blog/structured-data-and-unstructured-data-explained/.
Everlaw. 2022. “Everlaw Launches AI-Based Clustering to Open a New World of Ediscovery Insights to Legal Teams.” June 8, 2022. https://www.prnewswire.com/news-releases/everlaw-launches-ai-based-clustering-to-open-a-new-world-of-ediscovery-insights-to-legal-teams-301563544.html.
IBM. 2023a. “What Is Artificial Intelligence (AI) ?” IBM. 2023. https://www.ibm.com/topics/artificial-intelligence.
IBM. 2023b. “What Is Machine Learning?.” IBM. 2023. https://www.ibm.com/topics/machine-learning.
Linkedin. 2023. “What Are the Most Effective Ways to Use Machine Learning for Pattern Identification in Unstructured Data?” Accessed November 15, 2023. https://www.linkedin.com/advice/0/what-most-effective-ways-use-machine-learning-pattern.
Marwaha, Avaneesh. 2017. “Seven Benefits of Artificial Intelligence for Law Firms.” Law Technology Today. July 13, 2017. https://www.lawtechnologytoday.org/2017/07/seven-benefits-artificial-intelligence-law-firms/.
McMicheal, Jonathan. 2023. “Artificial Intelligence and the Research Paper: A Librarian’s Perspective.” SMU Libraries (blog). January 20, 2023. https://blog.smu.edu/smulibraries/2023/01/20/artificial-intelligence-and-the-research-paper-a-librarians-perspective/.
Reuters, Thomas. 2023. “Legal Research: 3-Step How-to Guide,” Thomas Reuters, October 2, 2023. https://legal.thomsonreuters.com/en/insights/articles/basics-of-legal-research-steps-to-follow.
Ryan, William, Allen Garrett, and Brad Sears. 2023. “Practical Lessons from the Attorney AI Missteps in Mata v. Avianca.” Association of Corporate Counsel. August 8, 2023. https://www.acc.com/resource-library/practical-lessons-attorney-ai-missteps-mata-v-avianca.
Silverman, Andy. 2019.“The Role of Machine Learning in Contract Analysis.” Contractworks. August 15, 2019. https://www.contractworks.com/blog/the-role-of-machine-learning-in-contract-analysis.
Strumberger, Sarah. 2023. “AI Contract Analysis Software: Buyer’s Guide for Legal Professionals.” Thomson Reuters Law Blog. September 5, 2023. https://legal.thomsonreuters.com/blog/buyers-guide-artificial-intelligence-in-contract-review-software/.
Welborn, Aaron. 2023. “ChatGPT and Fake Citations.” Duke University Libraries Blogs (blog). March 9, 2023. https://blogs.library.duke.edu/blog/2023/03/09/chatgpt-and-fake-citations/.