North Carolina’s 2026 budget law gave public school units a deadline: adopt policies governing educational AI use by June 30, 2027. It also assigned the state work on evaluating tools and supporting procurement. The institution must adopt a policy, but it does not have to build every part of the evaluation process alone.
That combination illustrates a question emerging across state laws. When a school board approves an AI policy, what has it actually put in place? A statement of acceptable use, a process for selecting products, and a system for overseeing their effects are different accomplishments.
This selective survey examines Arkansas, Ohio, Tennessee, Maryland, Virginia, and North Carolina as of October 16, 2026. It illustrates approaches to mandatory AI policies and related oversight, rather than providing an exhaustive inventory. Institutional coverage, deadlines, and implementation stages differ. Most examples concern public K–12 education; some expressly reach public higher education.
The Common Starting Point Is an Institutional Policy
Across these six examples, a recurring requirement is for the institution or its governing body to adopt an AI policy. The policy brings choices about permitted uses into an institutional process rather than leaving every teacher, department, or purchaser to decide independently.
Ohio’s HB96 requirement illustrates that starting point. Traditional public school districts, community schools, and STEM schools had to adopt policies by July 1, 2026. They could adopt the Department of Education and Workforce’s model or develop their own.
Ohio’s supporting materials discuss student and staff uses, privacy, vendor evaluation, and effects on learning and assessment. Those recommendations should be distinguished from the statute’s requirement to adopt a policy. The existence of a model does not make every recommended provision an independently binding statutory duty.
Privacy, staff preparation, acceptable uses, and human involvement recur across the selected approaches. They are not identical requirements in every state. A sound comparison identifies whether an element appears in the statute, required state guidance, or an optional model policy.
The distinction matters when an institution asks whether its policy complies. Following a useful recommendation is different from satisfying the particular mandate that applies to that institution.
Tennessee Makes Adoption Visible Beyond the Institution
Tennessee’s Public Chapter 550 requires policies on instructional and assignment uses by students and personnel. It expressly covers local school boards, public charter schools, the University of Tennessee trustees, the Board of Regents, and state-university governing boards.
The law was approved March 11, 2024. K–12 implementation began with the 2024–25 school year; the higher-education deadline was July 1, 2025. Its earlier enactment remains relevant to the implementation period examined here.
Local boards and charter governing bodies must report compliance annually, including their policies and how they will enforce them. The higher-education provisions require publication on the institution’s website and submission to the chairs of the senate education committee and the house education administration committee, with an appearance before the joint government operations committee for specified noncompliance.
The enhancement is visibility and follow-through. A policy does not remain solely an internal document. Reporting how it will be enforced asks the governing body to connect the adopted language to institutional action.
Arkansas Expressly Preserves Human Final Decisions
Arkansas Act 848 of 2025 applies through its public-entity framework, including public school districts, charter schools, and institutions of higher education. It requires policies defining authorized AI and automated-decision-tool uses and requires an authorized human employee or designee to make final decisions in the course of their employment.
The act also addresses employee training and disciplinary procedures for policy violations, and directs the Department of Education, in coordination with the State Cybersecurity Office, to develop policies for state educational institutions.
This supplies a clear rule about who makes the decision. It does not, by itself, establish what that person must know to evaluate a recommendation.
Consider a hypothetical administrator reviewing an AI-generated recommendation affecting a student. The administrator can formally approve the result without understanding the underlying information or limitations. Meaningful review requires a practical opportunity to identify a problem and change the outcome. That is an implementation question the institution must resolve; it is not a claim that the Arkansas act prescribes a particular review protocol.
Maryland Assigns a Coordinator and Requires Evaluation Tools
Maryland’s Artificial Intelligence Ready Schools Act, SB720, Chapter 634, took effect June 1, 2026. It requires local school systems to have policies aligned with departmental guidance within 120 days after that guidance is released. The effective date and the triggered policy deadline are distinct. A separate section of the act, establishing a State AI education collaborative, is time-limited and abrogates on May 31, 2028.
Each system must designate a central-office, noninstructional AI coordinator. The Department must publish a rubric and other evaluative tools, and local systems must use them when reviewing and selecting AI tools.
The enhancement connects responsibility to a repeatable evaluation process. Naming a coordinator identifies a person who can connect local implementation with the state. Requiring an evaluation rubric gives purchasers a common basis for examining products.
Neither step establishes that a particular product works. The practical question becomes whether the evidence supplied answers the rubric’s questions and supports the proposed educational use.
Virginia Connects Local Policies to Operating Oversight
Virginia’s section 22.1-20.2:1, enacted through HB1186 and SB394 in 2026, requires school boards to establish, implement, and enforce policies consistent with departmental guidance.
The statute specifies guidance addressing privacy, transparency, bias, accessibility, training, restrictions on external model training with school data, and prohibitions on relying solely on AI for certain high-stakes decisions defined by the Department. It also addresses teacher access to interaction transcripts and monitoring through dashboards, alerts, and audit logs.
This approach concerns what staff can see after a tool enters the classroom. A teacher cannot investigate an interaction the institution has no way to retrieve. An alert is useful only if someone receives it and knows what action to take.
Those practical questions also create privacy and workload considerations. More monitoring is not automatically better oversight. The institution needs to determine which information is necessary, who can access it, and how it will be used.
North Carolina Builds Shared Evaluation and Procurement Support
Section 7.39 of North Carolina Session Law 2026-41 requires a state model policy by December 31, 2026, access for public school units by January 15, 2027, and local adoption by June 30, 2027. The local policies follow review of the model; this should not be characterized as mandatory verbatim adoption.
The law also requires an evaluation framework that must address at least student data privacy, security, and transparency; alignment with the standard course of study; and accessibility for all students. It assigns annual framework review, public lists of reviewed tools and tools in use, and procurement support. It separately requires covered teachers to complete specified professional development by June 30, 2028.
These are enacted responsibilities with phased implementation. Their presence in the statute does not establish that every resource or list is already operational.
The enhancement is shared capacity. State evaluation resources can reduce duplicated work across districts. A reviewed-tool list can help identify relevant evidence, but it should not be treated as proof that every installation, population, or classroom use is appropriate.
