AI Education

State of play.

• Law school AI policy has fractured into competing institutional models, with UChicago's "AI-resilient" device-ban curriculum and Berkeley's default prohibition representing opposite ends of a spectrum now documented across 128 schools in Suffolk Dean Andrew Perlman's public archive .
• Norway has enacted a near-total generative AI ban for students ages 6-13, effective August 2026, citing cognitive-offloading concerns and supported by a study of 400 schools showing improved grades and a 60% reduction in school counseling visits—the first national-level age-tiered restriction and a potential template for other jurisdictions .
• The Florida Bar has made the first state bar move to institutionalize AI competency, partnering with Clio to offer free AI tool access and a certification path in responsible legal AI use to its members—a model other bars are watching .
• The Canvas/Instructure breach remains a live exposure, with class action investigations underway and notification obligations across thousands of institutions still dependent on whether ShinyHunters executed the threatened data release .
• For counsel advising educational institutions, law firms, or ed-tech clients, the picture has shifted from breach triage toward a longer-term structural question: what AI governance and training obligations are crystallizing, and which institutional models—restrictive, permissive, or hybrid—will define the standard of care when AI-related harms are litigated.

Where things stand.

• Law school AI policy is fragmented and rapidly evolving. Perlman's archive documents 279 policy records across 128 of 196 U.S. law schools, with significant variation: many prohibit generative AI in submitted work absent explicit authorization, few require written course-level disclosure, and the dataset remains incomplete as schools continue adopting new rules .
• The UChicago/Berkeley split is the national template moment. UChicago has moved toward supervised AI integration—banning devices in 1L core classes while embedding AI into research and writing courses and clinics; Berkeley has adopted a default prohibition covering conceptualization, outlining, drafting, revision, translation, and editing, with narrow exceptions . Firms hiring from both schools should expect material variation in associate AI competency and judgment.
• Norway's age-tiered restriction is the first national-level regulatory model for generative AI in K-12. The tiered framework—total ban for grades 1-7, supervised use for grades 8-10, encouraged literacy for upper secondary—distinguishes by developmental stage rather than tool type, and may influence how other jurisdictions structure liability frameworks around AI deployment in educational settings .
• The Florida Bar's Clio partnership is the first state bar AI competency initiative. The program offers four months of full Clio Work access plus a structured certification path; Clio's concurrent Legal AI Accelerator targets training 25,000 legal professionals . Whether other state bars replicate the model will determine whether bar-level AI competency becomes a national standard or remains episodic.
• Medical education has moved from pilot programs to mandatory AI integration frameworks. Major conferences including the Innovations in Medical Education Conference and HIMSS have issued concrete guidelines; Stanford's AI+Education Summit challenged assumptions about student assessment and AI literacy . State medical boards and accrediting bodies have not yet formally responded.
• AI-first private schools are expanding without independent validation. Alpha School has grown to more than 15 campuses with tuition of $40,000-$75,000 annually; its efficacy claims rest on proprietary internal analyses, multiple state regulators have rejected charter applications, and one campus serving low-income families experienced an 85% student departure rate .
• Single-vendor dependency remains the structural vulnerability the Canvas breach exposed. Canvas serves approximately 41% of U.S. educational institutions; ShinyHunters exploited Free-for-Teacher accounts to exfiltrate 3.65 terabytes of data, and a recompromise five days after initial containment raises questions about remediation adequacy .
• The Department of Education's Earnings Premium proposed rule would strip Direct Loan eligibility from programs whose graduates fail to out-earn high school graduates, with approximately 6% of programs—concentrated in for-profit institutions and low-wage certificate fields—facing material aid-eligibility risk .
• Corporate AI adoption bottleneck has shifted from deployment to workforce training. Asana's research found only 24% of marketers feel adequately trained in AI; companies are deploying internal champion networks rather than top-down mandates, raising informal-channel liability questions around bias, accuracy, and accountability .

Latest developments.

• Suffolk Dean Andrew Perlman launches public archive cataloging AI policies at 128 law schools, documenting 279 records across eight subject areas with significant variation in prohibition scope and disclosure requirements .
• UChicago and Berkeley adopt competing AI policies for the 2026 academic cycle—UChicago's supervised-integration model versus Berkeley's default prohibition—creating a national template split .
• Norway implements near-total generative AI ban for grades 1-7, effective August 2026, with tiered restrictions through upper secondary; the policy follows a study of 400 schools documenting improved outcomes under similar restrictions .
• Florida Bar announces Preferred Bar Program with Clio, making Florida the first state bar to offer complimentary legal AI access and a certification path in responsible use as a member benefit .
• Major medical education conferences declare AI integration mandatory, with the Innovations in Medical Education Conference, HIMSS, and Stanford's AI+Education Summit issuing concrete implementation guidelines rather than theoretical frameworks .
• Alpha School expands to 15-plus campuses with unverified efficacy claims, charter rejections by multiple state regulators, and an 85% student departure rate at its Brownsville campus .
• Companies deploy internal AI champion networks to address adoption gaps, with Asana research documenting only 24% of marketers feeling adequately trained—raising informal-channel liability questions for corporate clients .

Active questions and open splits.

• Which law school AI policy model will define the standard of care. The UChicago integration model and the Berkeley prohibition model are now both in operation at elite institutions; when AI-assisted work by a junior associate causes client harm, the question of what training the supervising firm should have provided will turn partly on what the associate's law school normalized .
• Whether Norway's age-tiered framework travels to U.S. jurisdictions. The Norwegian model is the first national regulatory precedent distinguishing by developmental stage; ed-tech companies, school districts, and AI developers should monitor whether state legislatures or the Department of Education adopt analogous frameworks—and what liability exposure attaches to deployment decisions made before such rules exist .
• Whether the Florida Bar model becomes a national AI competency standard. If other state bars replicate the Clio partnership structure, bar-level AI certification could evolve from a voluntary benefit into a component of competency obligations under professional responsibility rules—a shift that would affect every practicing attorney .
• AI-first school liability for unverified efficacy claims. Alpha School's expansion on the basis of proprietary, unvalidated performance data, combined with charter rejections and high attrition at low-income campuses, creates a consumer protection and false advertising exposure that state AGs have not yet formally pursued .
• Whether ShinyHunters executed the threatened Canvas data release. The May 12 deadline has passed; whether the attackers released data determines the scope and timing of notification obligations across thousands of institutions and the pace of class action filings against Instructure .
• Earnings Premium benchmark methodology and APA vulnerability. The proposed rule's use of high-school-graduate earnings as the benchmark for programs in low-wage fields like early childhood education may be vulnerable to arbitrary-and-capricious challenge or disparate-impact arguments; the final rule's methodology will determine the litigation landscape .
• What AI governance documentation satisfies an emerging institutional duty of care. As medical education conferences, state bars, and international regulators issue concrete frameworks, the question of what oversight protocols an educational institution must maintain—and what gap creates liability when AI-related harms materialize—remains unresolved .

What to watch.

• Whether other state bars announce partnerships or certification programs modeled on the Florida Bar-Clio initiative—and whether any bar formally incorporates AI competency into professional responsibility guidance.
• Whether U.S. state legislatures or the Department of Education propose age-tiered AI restrictions analogous to Norway's framework, and how ed-tech companies respond to the regulatory risk.
• First class action complaints against Instructure and whether institutional defendants are named; the pleading theories will define the edtech vendor liability landscape.
• Whether state AGs open consumer protection investigations into AI-first schools like Alpha School over unverified efficacy claims and equity concerns.
• Whether the Department of Education finalizes the Earnings Premium rule as proposed or modifies the benchmark methodology in response to comments—and whether higher education associations mount an APA challenge.
• How law firms adapt onboarding and supervision protocols as the first cohorts trained under the UChicago and Berkeley divergent AI policies enter practice.