What is a “Generative AI policy” supposed to do?
A guest blog post from Professor Michael J. Madison, University of Pittsburgh School of Law, Faculty Director of the Future Law Project and a John E. Murray Faculty Scholar
I want to focus attention on an under-appreciated dimension of the conversations and debates happening in and across both legal education and higher education when it comes to “what should we do about Generative AI?” That is: who is making these decisions, and on whose behalf, and how?
As the Fall 2026 semester gets underway, the waters of US and Canadian legal education are rife with institutional policies and guidance of related sorts directed partly to Generative AI use by students and partly to adjacent questions of student use of technology generally. Some of those, such as UC Berkeley’s mostly anti-AI policy and the University of Chciago’s laptop ban, have come in for a kicking here on AI for Legal Education. If you are keeping track of these things, the University of Toronto’s law program is also trying to limit student use of classroom technology; Columbia Law School came out with what it characterizes as a “balanced” approach to student experience with Generative AI. Among the most recent notable interventions in this flow of news is one from the University of Georgia’s School of Law, which sets a table of opportunity from which students should design their own paths.
Meanwhile, the broader university would like a say. The social sciences faculty at the University of Chicago, for example. And MIT has come out with a university-scale report.
Others can and have and will pick apart what they like and don’t like in each approach. There is (as well) a cottage industry of individual faculty members laying out their own approaches, which range from outright hostility (Brad Wendel (Cornell)) to grudging, skeptical acknowledgment (Michael Plaxton (Saskatchewan)) to acceptance coupled with pushing students to be purposeful and reflective about what they are doing (me) to “dogfoodism” (practice what you preach in AI enthusiasm) (Seth Chandler).
Here is what interests me, and my response to my introductory question. In what respects are schools making decisions about AI exposure, awareness, and training on behalf of their students? In what respects are individual faculty members making those decisions? In what respects are students being left to – for career development purposes – fend for themselves in figuring out what to learn about AI, and from whom, and how?
The institutional policies are, themselves, almost entirely matters of signaling rather than actual practice, and to the extent that the policies go beyond signaling, they represent public resource commitments by the schools and, in some cases, by their parent universities. Berkeley clearly and purposely wants to come across to various audiences, including to members of the Berkeley Law community, as “tech pragmatists,” if not “tech skeptics”; Berkeley’s law school has long been celebrated in other respects for its early and deep commitment to law-meets-tech research and teaching. Berkeley is not going to go out of its way to invest extra resources in AI-themed training. Across the San Francisco Bay, UC Law San Francisco has staked out the opposite pole, celebrating its commitment to having each and every student complete a capstone “AI enabled” lab. The University of Chicago Law School already employs a cadre of experienced and sophisticated legaltech practitioners in its ranks of Lecturers; I read the school’s AI policy as doubling down on commitments to have those faculty members steer the school’s AI future. The University of Georgia School of Law is levelling up, in a way; the school as such is currently not putting its shoulder into the AI space but is more than willing to support faculty colleagues who are actively synching up their efforts with AI-related resources at the University of Georgia, full stop.
At the margin, institutional policies may influence decision-making by students who opt to attend one law school rather than another based on their respective AI policies (I would be surprised if this happens much, if it happens at all); at the margin, the policies may affect legal employers’ choices to recruit at one law school rather than another, or to prefer candidates from one school over another, if there are choices to be made (again, I would be surprised if this happens much, if it happens at all).
In an important way, individual policies matter more than institutional policies. Individual teachers are almost always given the power (or retain the power) to design their own policies – based on principles of academic freedom as those meet the pragmatics of policy enforcement – even in environments otherwise described as “no AI for students,” institutionally. For obvious reasons, beginning with the fact that many if not most law professors, like many if not most university faculty generally, do not publicize the details of their syllabi or their course policies, it is difficult to get a sense of the scale of the variation that describes individual policies. I suspect that my micro-survey above of individual professors’ self-description offers only the tip of a large iceberg when it comes to what students will hear (or at times, not hear) from their teachers about when and how AI use is (acceptable) (not acceptable) (a great idea) (and so on).
All of which leads students to sort out an enormous amount of the “what to do about Generative AI?” question on their own. And without any particularly great or even good guidance either focused on Generative AI specifically or on broader questions of the roles and purposes of technology in professional practice generally, and how those might bear on decisions about investing in curricular, co-curricular, and extra-curricular programs.
Looking at the issue from one perspective – that of the “classic” law school curriculum – it might be difficult to see a problem. Law students have long been tasked with sorting out curricular choices and part-time and summer employment largely on their own, taking cues from upper-level students, the occasionally helpful faculty mentor, a little nudging from a career services office staff, and general knowledge about what it means to (to learn to) be a lawyer. What’s one more complex body of information and prospective practice that our students should have to know something about?
Computerized legal research was a skill to be encountered and perhaps mastered, but no one ever had to be guided in selecting among normative frames for appreciating the place of computerized legal research in lawyer formation or legal practice. No one ever said – meaningfully – that Westlaw would corrupt a law student’s capacity for critical thought or expression.
My point, then, is that in coming up with policies – whether as institutions or as individuals – we are individually and collectively failing to appreciate the fact that we are imposing a significant decision-making burden on our students.
As a sector, legal educators spend essentially zero time as it is instructing our students in decision-making.
The school says “X” about Generative AI; Professor A says, “I have good reasons to promote and teach “Not X”; Professor B says, “I promote and teach Y about Generative AI.” Beyond the walls of the law school, loud voices decry the concentration of power in a handful of frontier AI firms in Silicon Valley, the labor market and environmental harms that many say accompany or will accompany the widespread deployment of AI hardware and systems, and the many ways in which data centers are corrupting both rural and suburban landscapes much as ChatGPT is corrupting the interpretive skills of students and the literary skills of higher education administrators.
“What it means to be a lawyer” used to be a something helpful practical principle in selecting among and prioritizing how to spend one’s time in law school. But (in a post for the future) the implicit “models of law practice” that have long shaped the design of the law school curriculum are losing much of their integrative force. “What it means to be a lawyer” today is now many different things – related, but no longer as clearly integrated as they were as recently as 25 or 30 years ago.
Even if one disagrees with that intuition, today’s practicing bar is little help. I was on a Zoom call earlier this week with several Chief Innovation Officers of large law firms, and the dominant theme in their descriptions of new hires, straight out of law school, was a complaint that new lawyers don’t know how much tokens cost. As usual, many if not most law firms expect a grounding in the details of legaltech use that is literally beyond the capacity of most law schools to deliver. (Again, how law schools are trying to manage that capacity issue is a post for a different day.)
Given the massive heterogeneity of law school and law professor policy and practice as to Generative AI, even within a given school, what should a student do? What should a cohort of students do? My question is not about the ultimate outcome. My question goes to how the decision should be made.
AI is not failing our students.
We are.