Skip to Main Content

ChatGPT and Generative AI Legal Research Guide

ChatGPT Performance on Bar Exams

Latest Version of ChatGPT Aces Bar Exam with Score Nearing 90th Percentile, ABA Journal, March 16, 2023

According to an article on ABA Journal, OpenAI’s ChatGPT-4 has achieved a score in the 90th percentile on the Multistate Bar Examination (MBE), which is used by many states as part of their bar exam. The article discusses how the use of AI in legal work raises ethical questions about the role of lawyers and the quality of legal work.

New GPT-4 Passes All Sections of The Uniform Bar Exam. Maybe This Will Finally Kill the Bar Exam, Above the Law, March 14, 2023

According to an article on Above the Law, OpenAI’s GPT-4 has passed all sections of the Uniform Bar Exam (UBE). The article discusses how AI can be used to help people study for exams and how AI can be used to help people learn new skills. The article also mentions that the use of AI raises ethical questions about the role of humans and the quality of work produced by AI.

Could AI Graduate from Law School? Minnesota Professors Put ChatGPT to the Test — and it Gets a C+, Minnesota Lawyers, February 22, 2023

This article describes an experiment conducted by two law professors from the University of Minnesota, who gave four law exams to ChatGPT, an AI chatbot based on GPT-3.5. The article reveals that ChatGPT passed all four exams with a C+ average, which is better than some human students. It also discusses some of the strengths and weaknesses of ChatGPT’s performance and its implications for legal education and practice.

ChatGPT Almost Passed the Bar, But Competent Lawyers Do Much More, Bloomberg Law, February 13, 2023

This article reports that ChatGPT, an AI chatbot based on GPT-3.5, scored 67% on a mock bar exam, which is close to the passing score of 70%. However, it also argues that passing the bar exam does not mean that ChatGPT can replace lawyers, who do much more than answering multiple-choice questions. It emphasizes the importance of human skills such as judgment, creativity and empathy in legal practice.

ChatGPT Passes Exams from Law and Business Schools, CNN, January 26, 2023

This article reports that ChatGPT, an AI chatbot based on GPT-3.5, passed five graduate-level exams from law and business schools. It says that ChatGPT scored a C+ average on four law exams from the University of Minnesota and a B- on a business exam from the University of Pennsylvania’s Wharton School. It also quotes some professors who conducted the tests and shared their insights on ChatGPT’s capabilities and limitations.

GPT Takes the Bar Exam, SSRN, January 3, 2023

Mike Bommarito and Dan Katz used ChatGPT-3.5, an AI language model developed by OpenAI, to tackle the Multistate Bar Examination (MBE) multiple-choice questions. Despite the baseline guessing rate for the exam being 25% for four-choice questions, ChatGPT-3.5 achieved an impressive overall accuracy rate of 50.3% without any specific training. This performance surpassed that of many law school graduates, with ChatGPT-3.5 even reaching the average passing rate for two subjects: Evidence and Torts. The success of this experiment raises questions about the potential of AI in the field of legal education and testing.

GPT Takes the Bar Exam at the 2023 MIT Computational Law Workshop, MIT Computational Law, January 19, 2023 (YouTube)

This video shows a presentation by Professor Dan Katz and Michael Bommarito on their research paper “GPT Takes the Bar Exam”. They demonstrate how they used ChatGPT-3.5, a large-scale language model, to answer multiple-choice questions from the Multistate Bar Examination (MBE). They report that ChatGPT-3.5 achieved a correct rate of 50.3% on a complete MBE practice exam, which is significantly higher than random guessing and close to passing rates for some subjects.