In a policy announced Thursday, and effective this summer, the law school will prohibit its 1,120 students from using artificial intelligence in preparing class assignments and in all exams — except for courses designed to teach students how to use AI, ethically and legally, while practicing law. Those courses are generally for students who are nearing graduation.
Exceptions will also be made for teachers who want to give their students more leeway in using the technology. But it will generally be barred for end-of-term exams.
Students will also be prohibited, the university announced, from using AI devices “for aid in conceptualizing, outlining, drafting, revising, translating, or editing any work submitted for credit.”
The current policy, in effect since 2024, restricts AI at the law school but allows students to use the technology for a few specified purposes, like correcting their grammar in research papers. Teachers say that students are still turning in papers citing cases the teachers had never heard of, evidently the product of AI.
Or a student asked to solve a legal problem will report that “I found the solution, in a chat with AI,” when “what really happened is that AI found the solution in existing (cases) and presented it to the student as a new idea,” said Chris Hoofnagle, a Berkeley law professor. “That’s actually plagiarism.”
AI use has been rapidly expanding in the legal profession, where a worldwide survey recently found that 80 percent of large law firms said they were using or exploring use of the technology. And its misuse has also been widespread — another report found nearly 1,400 cases, including 957 in the United States, in which courts have found that artificial intelligence was used to cite “hallucinated” cases — cases that were misrepresented, or nonexistent — in legal filings. Legal commentators said the actual number is almost certainly much higher.
Hoofnagle, who teaches first-year law students at UC Berkeley, said he proposed the new rules to other faculty members after seeing increasing numbers of assigned classwork with questionable legal reasoning.
The typical problem in law school, he said, has not been hallucinated cases, but defective legal analysis of actual cases. One apparent reason, he said, is that the policy in effect since 2024 has allowed students to consult AI technology for “brainstorming,” or coming up with ideas for class assignments and future exams.
“If you don’t have your own analytical judgment, AI will do it for you, and then it’s no longer your judgment,” Hoofnagle told the Chronicle. “It’s OpenAI or Anthropic,” two of the leading artificial intelligence companies.
He said Stanford Law School and several institutions elsewhere have similar policies.
The banned practices at UC Berkeley will include asking AI to propose a topic for a class assignment; asking for a summary of a legal rule for use in an assignment; seeking help in identifying repetitive passages in a student’s work or correcting grammar, and requesting an outline the student might use in a future exam.
AI will still be allowed for tutoring students, and for study guides, but not for class assignments, Hoofnagle said.
Another Berkeley law professor, Jonathan Glater, said the policy seems flexible enough to let him show students how to use AI, effectively as well as ethically, while preventing improper reliance.
“There are ways to use the technology that can up a lawyer’s game, and if I am teaching a more advanced class that aims to prepare them to use this tool in practice, it would be weird to prohibit its use,” Glater said by email.
But he said the strict ban “makes a lot of sense … in an introductory class that is helping students learn basic and essential concepts” for studying and learning law.
© 2026 the San Francisco Chronicle. Distributed by Tribune Content Agency, LLC.
