The UC Berkeley Law School is tightening its regulations regarding the use of artificial intelligence in academic settings. While the institution does not intend to impose a complete ban on AI technologies, the revised policy restricts students from using AI tools in various contexts, including brainstorming sessions. Professor Chris Hoofnagle, a key contributor to the policy, explained that the goal is to equip students with essential skills needed for practicing law in an AI-centric landscape.
According to Hoofnagle, the university's 2023 policy was deemed "too liberal" in how it permitted AI usage among students, especially in light of recent advancements in generative AI technologies that can autonomously create complete research papers. "As the capabilities of large language models (LLMs) have grown, we needed to reassess how much students depend on these tools," he stated.
The updated policy, set to take effect this summer, prohibits students from utilizing AI for various academic tasks, such as brainstorming, outlining, drafting, revising, and translating, especially during exams. In contrast, the earlier policy had allowed AI-assisted brainstorming, including using chatbots to generate paper topics.
The new guidelines received approval through a faculty vote, yet they allow instructors latitude to make exceptions. Additionally, courses centered around AI may implement different standards. The overarching intention is to ensure first-year law students grasp the foundational aspects of legal practice, including the skills to read, analyze, and articulate case law effectively.
Hoofnagle emphasized the importance of developing analytical judgment among law students. "The core question is what value a lawyer brings," he explained. "If an attorney cannot critically evaluate AI-generated outputs, their role diminishes significantly. Our policy aims to address this concern."
He further acknowledged the expectation from law firms for graduates to have practical skills in using AI technologies. "Students are increasingly requesting these courses, and they discover during internships that many firms already integrate AI extensively into their operations," he noted.
The legal tech landscape is rapidly evolving, with startups like Harvey and Legora vying for a slice of the projected $1 trillion global legal market. Harvey aims to enhance its presence in legal education by providing complimentary access to law school resources. In comparison, Stanford Law School, which Hoofnagle identified as having a more stringent AI policy during Berkeley's initial rollout, participates in Harvey's alliance program for law schools.
Despite the new policy, Hoofnagle admitted there are inherent challenges and loopholes. He pointed out that managing AI use is increasingly complex, particularly as search engines integrate AI technologies to remain relevant. Even research tools like Lexis and Westlaw now feature LLM-generated summaries, complicating efforts to enforce these policies. "There isn't a straightforward solution to this issue," he admitted, acknowledging that institutions cannot prohibit all forms of search.
The rapid evolution of AI is prompting both law schools and universities to adapt swiftly to these changes and the issues they bring. For example, Princeton University has recently updated its honor code for the first time in 133 years, introducing mandatory proctoring for all in-person exams starting July 1. This decision was partially driven by concerns surrounding the rise of AI.
At Berkeley Law, Hoofnagle noted an increase in misconduct cases and the shift of more take-home exams to in-person formats. The in-person assessments utilize specialized software designed to prevent internet access and disable copy-and-paste functions. Nonetheless, he cautioned that cheating can still occur. "While it's challenging to safeguard against every risk, there's a strong message that cutting corners can ultimately backfire when students face in-person testing," he explained.
