Columbia Law School has released a new artificial intelligence policy ahead of the school year’s autumn start, telling students it supports them using the new technology to research, learn and understand legal concepts.
But the school made clear students are not to write their legal analysis, form arguments, or take exams — unless they tell their professors they employed it in their studies.
In the new guidance, Dean Daniel Abebe was clear that Columbia Law is not endorsing “unrestricted use” of AI. Instead, the school said it wanted to encourage students to think critically about the new tech and learn how to use it to their advantage so all enrolled would be ready to mesh with the way the firms they’d graduate into would inevitably be using the technology when they got out of school.
The policy seemed to spring from a sense that AI wasn’t going away – and is already being incorporated into legal practice – so there’s no reason to fight it.
“As AI transforms the study and practice of law, the challenge before us is not whether to engage with AI, but how to do so in a way that advances the Law School’s academic and educational purpose: teaching our students how to think critically and rigorously, and helping them develop the judgment, discipline, and technological fluency that modern legal practice increasingly demands,” Abebe wrote in a statement, adding that he believed the policy “reflected” these commitments.
Students will bear “full intellectual responsibility” for all submitted work — whether they use AI to produce it or not — and the work they turn in will be taken as reflecting their judgment, reasoning, and analysis.
Notably, the policy permits students to use AI to help them learn and understand legal concepts – doing things like asking the tech to summarize their own notes, generate examples or hypothetical situations, define legal terms, study, and even serve as a mock argument partner – as well as refine and improve the clarity of their writing and correct spelling and grammar errors.
Jerome L. Greene Hall at Columbia Law School, where hundreds of students will return this fall with AI policies that allow them to use the tool in their education.Provided by Columbia Law School
It also permits students to use AI for research, like asking for help identifying potentially relevant cases and summarizing existing case law — something that got attorneys in trouble in a growing number of courts — including in the Southern District of New York – for putting forth AI-hallucinated, fake cases as valid legal precedent for their arguments.
The policy adds that “in all of these cases, students are required to independently and rigorously assess the feedback generated by AI, using their own critical judgment to decide whether and how to implement any suggestions,” and they must tell their professors anytime they use the tool.
While a student can use an AI tool to “refine” their writing, Columbia bars the use of AI for the actual generation of that writing of legal arguments and analysis. The policy clearly states that submitted work must be purely student-composed, with ideas born of “their own intellectual creativity and judgment,” and cautions that simply revising AI-generated outputs would not fly.
“At every stage of the writing process, the ideas, arguments, analysis, and expression in submitted work must reflect the student’s own judgment and thinking,” the policy states. “Students must deliberately choose the arguments they advance, understand and be able to explain and defend every proposition in their submissions, and accept responsibility for their accuracy and conclusions.”
It also prohibits students from using AI in exams, participating or being present in class. Earlier this year, a plaintiff was chewed out by a New York appellate judge because they had presented filings from an AI-generated attorney in the courtroom.
The new policy also bars students from placing any confidential information into an AI system when they are producing work for actual people, or for use in a classroom setting. They also may not use AI for an externship for credit.
“AI may serve as a learning aid and critical interlocutor—helping students test ideas, improve organization, or refine expression,” the policy says. “AI may not, however, replace the student’s own legal reasoning, serve as an undisclosed ghostwriter, or otherwise replace the student’s role in originating the intellectual content of a submission.”
“A student who submits work generated by AI without exercising independent judgment, or who cannot explain and defend the work as their own, has failed to satisfy the expectations,” the policy adds.
The school said the policy was developed based on recommendations by its AI Task Force that began work in the fall of 2025 and reflects input from Law School faculty, students, administrators, and alumni. Professors will be allowed to adopt standards specific to their classrooms as they see fit.
“After consulting widely among faculty, students, alumni, and staff, the Task Force coalesced around the idea that students should be equipped with the adaptive conceptual frameworks necessary to navigate a shifting landscape, rather than merely trained on specific tools that may be obsolete by the time they graduate,” the policy reads. “As technological fluency becomes an essential component of legal practice, the effective and ethical use of AI will depend on the quality of the thinking among those who guide its use.”
The law school said it plans to revise its policy as the technology continues to develop and that its “commitment to a rigorous legal education remains steadfast.”
“We recognize that the technological landscape is rapidly evolving, and we are committed to an iterative and thoughtful process of review and refinement,” the policy states. “This policy seeks to advance core institutional values—fostering critical thinking and independent judgment—while carefully integrating the AI that are becoming essential to the practice of law.”
