Table of Contents
Law School AI Policy Split: Ban Laptops or Require AI?
At least 12 schools changed law school AI policy this summer, and 36 of 180 now require AI instruction. Why both camps are right, and what K-12 copies next.
American law schools spent the summer of 2026 going in two directions at once. By September 21, 2026, at least 12 U.S. law schools had changed their law school AI policy, and at least 36 of 180 surveyed now require AI instruction, while others moved the opposite way and restricted laptops in class to protect the Socratic method. That is not confusion. It is two coherent strategies optimising for two different outcomes, and the same fork is arriving in high schools. A parent watching this fight is getting a free preview of the argument their own school board will have in about two years.
Key Takeaways
- At least 12 law schools changed AI policy over one summer. At least 36 of 180 surveyed now require AI instruction, roughly one in five.
- The restriction camp is not mainly worried about cheating. It is protecting a specific cognitive practice: reasoning aloud, unassisted, under pressure.
- The requirement camp is not mainly excited about technology. It is responding to the fact that graduates will be expected to use these tools competently on day one.
- These are school-level decisions, not a national mandate. No accreditation rule forced any of them, which is why the picture looks scattered.
- The two strategies are compatible if you separate them by task. Restrict during the practice that builds the skill; require during the work that applies it.
What changed, and at what scale
An institution-level AI policy is a rule set by a single school about what tools students may use in which contexts, enforced through its own academic rules.
The AI-in-education policy tracker entry dated September 21, 2026 reports that at least 12 U.S. law schools changed AI policies over the summer, that at least 36 of 180 surveyed now require AI instruction, and that some schools restrict laptops to protect the Socratic method.
Be careful with those numbers in both directions. “At least 12” is a floor from a survey, not a census, so the real figure is higher. “36 of 180” is a meaningful denominator and tells you the requirement camp is a substantial minority rather than a trend that has already won. And nothing in that entry describes an accreditation requirement. These are individual faculties voting, which is why the map looks patchy and why it will keep moving.
Why ban laptops in a professional school in 2026
This is the part most coverage gets wrong, so it is worth the mechanics.
The Socratic method, as law schools use it, is a live interrogation. The professor asks a student to state a case, then pushes: what if the facts changed, what is the rule, why does the rule not apply here. The pedagogical work happens in the gap between the question and the answer, while the student is generating a response from memory and reasoning rather than retrieving it from a document.
That gap is doing something specific. UCLA’s Bjork Learning and Forgetting Lab describes a set of desirable difficulties, “difficult but effective training conditions” that impede immediate performance while producing better long-term retention. Two of them are directly at stake here: generation, producing an answer rather than reading one, and testing, retrieving information rather than restudying it. The lab’s summary is blunt: “performance during training is not always representative of long-term learning.”
A laptop with a chatbot open removes both. The student can produce a better answer in the moment and learn less from producing it. That is the same signature faculty elsewhere are describing as an illusion of learning, and it is the reason a laptop restriction in a Socratic classroom is a pedagogical decision rather than a disciplinary one. It is not about catching anyone. It is about preserving a condition under which the exercise works.
Why require AI instruction in the same building
The other camp is answering a different question: what does a graduate need to be able to do?
Legal work involves document review, drafting, research and summarising at volume, and firms have adopted AI tools for all four. A graduate who cannot supervise a model’s output, spot a fabricated citation, or explain to a client what a tool did with their file is less employable, and arguably less safe. Requiring instruction is a competence argument, not an enthusiasm argument.
There is also a professional-responsibility dimension that makes law unusual. A lawyer who files a brief containing invented case law has a problem that goes beyond a bad grade. Teaching students how these systems fail is risk management for the profession, which is why “require AI instruction” and “ban it from the Socratic classroom” can sit in the same course catalogue without contradiction.
| Approach | What it protects | What it costs | The K-12 analogue |
|---|---|---|---|
| Restrict devices in class | Generation and retrieval, the conditions that build durable skill | Note-taking convenience; students with accommodations need carve-outs | In-class writing, phone-free math, oral explanation |
| Require AI instruction | Graduate competence and professional safety | Faculty time; curriculum space taken from something else | A required AI literacy unit, taught rather than assumed |
| Both, separated by task | Skill building and skill application | Coordination; a clear rule per assignment | ”Tool-free practice, tool-allowed production” |
| Neither, left to instructors | Flexibility | Students get contradictory rules across four classes | The most common district posture today |
That bottom row is where most American schools, K–12 included, actually sit. Ask your child how many different AI rules they are operating under this term. If the answer is more than two, the inconsistency is itself the policy.
What the broader evidence says about this fork
The experimental literature supports the restrict-then-require sequence more cleanly than either camp alone.
Bastani and colleagues, publishing in PNAS in 2025, found high school students with GPT-4 tutoring performed better during practice and showed reduced skill development once access was removed, with the harm mitigated when the tutor gave teacher-designed hints instead of answers. That is an argument for restricting assistance during the skill-building phase.
Kestin and colleagues, publishing in Scientific Reports in 2025, found college physics students using a purpose-built AI tutor “learn significantly more in less time” than peers in in-class active learning. That is an argument for teaching students to use well-designed tools well.
Michael Gerlich’s 2025 survey of 666 UK participants in Societies found a strong negative association between AI tool use and critical thinking (r = −0.68), mediated by cognitive offloading. It is correlational and about adults, so treat it as consistent with the mechanism rather than as proof of it.
None of these studies tested a laptop ban. Anyone citing research as direct support for a device restriction is overreaching. What the research supports is the underlying claim: assistance during practice inflates performance and can deflate learning, and design determines which happens.
What to do at home
Find out your child’s rule per class, not per school
Districts publish one policy; teachers implement four. Ask your child to write down, for each class, whether AI is allowed for brainstorming, for drafting, for checking, and for finishing. The pattern of blanks is the conversation to have with the school, and it is far more concrete than “what is your AI policy.”
Adopt the restrict-then-require sequence at home
Practice with the tool closed, produce with the tool open, and be explicit about which phase you are in. Saying “this is practice, so nothing open” is a clearer rule than a general ban, and children follow phase rules better than blanket ones because the reasoning is visible.
Train supervision, not avoidance
The law schools requiring AI instruction are teaching students to catch a model’s errors. Do the same thing with a 13-year-old: ask the tool for a short summary of something your child already knows well, then have them mark what it got wrong. Finding the error is the skill. Our guide to teaching kids AI ethics covers the judgement side of this.
Expect oral and in-class formats to spread
If law schools are protecting live reasoning, high schools will too, and the format is coming to your child’s classroom whether or not the AI policy changes. Practising spoken explanation at the dinner table is low-cost preparation for an assessment shift that is already underway.
What not to do
Do not treat the laptop restriction as technophobia. It is the same logic as closing the book before a retell, applied to a professional school. And do not treat mandatory AI courses as capitulation. The useful critique of both camps is not that they are wrong, it is that a school adopting only one of them has answered only half the question. Our overview of whether schools are preparing students for an AI economy covers the other half.
What to Watch For Over the Next 3 Months
- Week 4: Watch for the number to move. “At least 12” and “36 of 180” are floors from a late-summer survey, and spring-semester catalogues will revise both upward.
- Month 2 red flags: A school that bans devices without an accommodations carve-out. A required AI course taught with no mention of failure modes. A high school citing law schools as justification for a policy that copies only the restriction half.
- Month 3 self-check: Can your child say which phase of an assignment they are in, and which tools are allowed in that phase? If the answer is a single rule for everything, the rule is probably being ignored.
Frequently Asked Questions
How many law schools have actually changed their AI policy?
At least 12 changed policies over the summer of 2026, and at least 36 of 180 surveyed now require AI instruction. Both figures are floors from a survey rather than a complete count, so the real numbers are higher.
Why would a law school ban laptops?
To protect the Socratic method, which depends on a student generating an answer from memory and reasoning under questioning. Generation and retrieval are among the training conditions UCLA’s Bjork lab identifies as producing better long-term retention, and an open device removes both.
Is requiring AI instruction a national rule?
No. The tracker describes school-level policy changes, not an accreditation mandate. Each faculty is deciding independently, which is why the landscape looks inconsistent.
Do these two approaches contradict each other?
Not if you separate them by task. Restricting assistance during skill-building practice and requiring competent tool use during applied work are complementary. A school doing only one has answered half the question.
Does research support banning devices in class?
Not directly. No study cited here tested a laptop ban. The research supports the underlying mechanism, that assistance during practice can inflate performance and reduce retained learning, which is a reason to restrict during practice specifically.
What should high schools copy from this?
The sequencing, not the ban. A clear per-assignment rule about which phase allows which tools, plus explicit instruction in how models fail, covers both camps’ concerns without forcing a school to pick a side.
About the author
Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.
Sources
- Pursuit. (2026). “AI in Education: News, Policies, Innovations” (entry dated September 21, 2026). https://www.pursuit.us/news/ai-in-education-news-policies-innovations
- Bjork Learning and Forgetting Lab, University of California, Los Angeles. “Research: Desirable Difficulties.” https://bjorklab.psych.ucla.edu/research/
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). “Generative AI without guardrails can harm learning: Evidence from high school mathematics.” Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2422633122
- Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). “AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting.” Scientific Reports. https://www.nature.com/articles/s41598-025-97652-6
- Gerlich, M. (2025). “AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking.” Societies, 15(1), 6. https://www.mdpi.com/2075-4698/15/1/6
- Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.” arXiv:2506.08872. https://arxiv.org/abs/2506.08872
- UNESCO. (2024). “AI Competency Framework for Students.” https://www.unesco.org/en/articles/ai-competency-framework-students
