Berkeley law school adopts sweeping restrictions on student AI use

UC Berkeley School of Law is drawing a sharp line against generative artificial intelligence, adopting one of the nation's most restrictive AI policies for law students at a moment when courts and law firms are rapidly embracing the technology.

Beginning this summer, Berkeley law students will be barred from using AI tools not only to draft papers, but also to brainstorm ideas, organize arguments, revise prose, correct grammar and translate written work into English. The policy also prohibits students from uploading class materials -- including readings, slides and recordings -- into generative AI systems.

The move reflects growing anxiety inside legal academia that students may become overly dependent on AI at the expense of developing the analytical reasoning and writing skills central to lawyering. But critics say Berkeley risks moving badly out of step with a profession already integrating AI into daily practice.

"Judges are using these tools. Law firms are using these tools," said Daniel W. Linna Jr., director of Law and Technology Initiatives at Northwestern Pritzker School of Law. "Junior people in law firms are using these tools. They're training people to use them responsibly."

Berkeley's new policy marks a major shift from its earlier rules, which focused primarily on plagiarism and permitted limited uses such as grammar correction and research support. Under the revised policy, AI may be used only for narrow research purposes, such as locating cases, statutes, and secondary sources. Students remain responsible for verifying all citations and research results.

The policy bans a broad array of common AI uses, including asking chatbots to brainstorm paper topics, create outlines, summarize legal rules, or identify repetitive passages for editing. AI use is forbidden in any exam setting and may not be used to translate papers written in another language into English.

Law school officials framed the policy as an effort to preserve independent thinking.

"Thinking remains the sine qua non of good lawyering (and of a quality legal education)," the policy states. "This policy seeks to ensure that our courses focus on requisite cognitive skills by default."

UC Berkeley law professor Chris Hoofnagle, faculty director of the Berkeley Center for Law and Technology, defended the approach in a recent post on X, where a debate has raged for days over the policy.

"In the classroom, we don't want students to write the best possible paper, but rather the best possible paper that the student is capable of," Hoofnagle wrote.

The policy allows professors to make exceptions for courses specifically designed to teach AI fluency, though any authorized AI use must be disclosed by students.

Linna called parts of the policy "extreme," particularly the ban on grammar correction and translation tools. He said the rules could disproportionately affect first-generation students, international students and non-native English speakers. He also questioned how the policy could be fairly enforced, pointing to concerns about unreliable AI-detection software and subjective faculty assumptions about AI-generated writing.

"It's not just that it cannot be enforced," Linna said. "It's going to be unfairly enforced."

More fundamentally, Linna argued that law schools should rethink pedagogy rather than attempt to wall off AI entirely. At Northwestern, faculty are exploring AI-powered simulations that allow students to practice oral arguments, negotiations and classroom cold calls.

"Undoubtedly, if we don't change the way we teach, and students use AI, we're not going to get good educational outcomes," Linna said. "But that doesn't mean we ban it."

William H.J. Hubbard, deputy dean at the University of Chicago Law School, said Chicago's current policy is far more permissive than Berkeley's. Chicago prohibits AI use during exams and bars plagiarism-style misuse, but does not prohibit grammar correction or limited assistance tools.

"Berkeley's new policy is much broader, certainly, than our policy," Hubbard said.

Still, Hubbard said Berkeley's concerns are legitimate. Law schools, he said, face a genuine challenge in preventing students from becoming too reliant on AI.

"It's possible that students who, without sufficient guidance, will use AI so much that they don't learn as much," Hubbard said. "That, I think, is a very legitimate concern."

Rather than broad bans, however, Chicago is considering structural changes to classrooms and exams that reduce opportunities for improper AI use. Some professors already prohibit laptops during classroom discussions to prevent students from consulting chatbots during cold calls.

"Rather than prohibiting AI use and then facing the very difficult challenge of enforcing the prohibition," Hubbard said, "instead, you're simply designing the classroom in a way that makes the AI use impossible."

Staff Writer Craig Anderson contributed to this report.