First-year law at schools like these has always been a particular kind of ordeal. Professors drill students with rapid-fire questions, forcing them to justify conclusions on the spot. The Socratic method depends on the uncomfortable silence, on the student who has to reason out loud. That pedagogical machinery does not work if the reasoning quietly happens inside a laptop first.
What changed this autumn
The University of Chicago Law School announced its pilot programme in July 2026: laptops, tablets and smartphones are banned in all core first-year (1L) courses for the entire 2026–27 academic year. The result is pen-and-paper note-taking in class and answers built from reading and argument, not from a search box.
UC Berkeley Law went further in scope. Its default policy bans AI use across all student coursework unless an instructor explicitly opts out. Where the default applies, Berkeley spells out six stages of writing that may not be handed to software: conceptualizing, outlining, drafting, revising, translating and editing. That list matters because it targets the entire pipeline of legal thinking, not just the final answer.
Columbia Law School's 2026–27 rules take yet another form: AI may not replace legal reasoning, and AI is barred from exams.
None of this is a simple anti-technology gesture. As Patti Waldmeir's opinion piece in the Financial Times points out, law faculties are trying to protect the analytical skills that make a lawyer a lawyer. The approach has a name: "AI-resilient pedagogy". It is two-track — restrict technology in foundational first-year courses, then teach controlled, responsible AI use in clinics and upper-year electives. Put the tool away first, learn to use it well later.
Ban the chatbot, but applications are booming
The timing is striking. Just as law schools lock AI out of the classroom, demand for legal education is surging. According to LSAC data from the latest admissions cycle, 69,555 people applied to law school — a 12% year-on-year increase and 30.2% above the four-year average. Those applicants submitted nearly 511,000 applications, up 14.6%.
Students, it seems, still believe legal training is worth the cost at a moment when machines can draft contracts and memoranda. That is exactly why the schools are tightening their pedagogy: if AI can produce competent boilerplate, the value of a law degree shifts toward judgment, argumentation, and the ability to spot what a model got wrong. Those are skills built through practice, not through prompting.
Europe: regulate, don't remove
Seen from Europe, the contrast could hardly be sharper. As of August 2026, the EU AI Act's general provisions are in full statutory enforcement. The transitional period of relying on voluntary draft codes of practice is over. National authorities and the EU AI Office now have active oversight and fining mechanisms, and the transparency obligations of Article 50 are mandatory.
Nothing in that framework tells a law faculty to ban ChatGPT. In fact, the obligations point in the opposite direction. Institutions that deploy AI tools must ensure adequate AI literacy among staff and students; transparency rules govern AI-generated content in specific contexts. The EU has chosen a model of obligations on providers and deployers, not classroom-level prohibition.
So there is no EU-wide equivalent to the Chicago laptop ban — and European law schools are unlikely to copy it. A European faculty reading about Berkeley's six prohibited writing stages will recognise the concern about critical thinking, but its regulatory homework lies elsewhere: teaching students what the AI Act demands when AI is used in professional legal work, and building habits of disclosure and oversight.
Two routes, same destination
Both approaches rest on a shared fear: that a lawyer who outsources the first draft to a chatbot never learns to detect a subtle flaw in its reasoning — or to catch a confident citation of a case that does not exist. The American answer is to isolate the student from the tool in the formative years. The European answer is to regulate the context in which the tool is used and make the user responsible.
It is too early to say which produces better lawyers. The honest test will come years from now, when today's first-year students appear in courtrooms — some trained to think first without machines, others trained from the start to use machines under clear legal obligations. What unites them is the recognition that legal reasoning remains the core product. The debate is only about the best way to protect it.
Does the University of Chicago ban mean law students cannot use AI anywhere?
No. The ban applies to electronic devices in all core first-year classes during 2026–27. Chicago's pilot targets classroom time, not studying at home. Berkeley's default ban is broader — it covers AI use in all coursework unless an instructor explicitly allows it — while Columbia's policy focuses on exams and on AI replacing legal reasoning.
Could European law schools introduce bans like these under the AI Act?
Nothing in the EU AI Act requires or encourages classroom electronics bans. The Act regulates providers and deployers of AI through transparency, risk management and AI literacy obligations. Academic integrity rules remain a matter for individual universities, and the general-purpose chatbots used by students are not prohibited by EU law.
Why teach AI use at all if it can weaken legal thinking?
Because legal practice already runs on AI-assisted research, document review and drafting tools. The US schools themselves take a two-track approach: restrict AI in foundational courses, then teach controlled, responsible use in clinics and advanced electives. The open question is not whether law students should learn AI — it is when and under what rules.
