Maryland Mandates a District AI Coordinator: K-12 AI Governance Moves From the Document to the Org Chart

Framing

For most of the past two years, the hardest question in K-12 AI governance has been who, inside a district, actually owns it. Policies were drafted by committees, signed by boards, and then left without a named office accountable for keeping them current as the tools changed weekly. On May 26, 2026, Maryland answered that question by statute. The Artificial Intelligence Ready Schools Act requires every local school system in the state to designate a non-instructional, central-office coordinator responsible for AI use, alongside a state obligation for the Maryland State Department of Education to publish guidance, best practices, and evaluative tools, and a new statewide AI Education Collaborative to study implementation and recommend policy. This is a different kind of governance signal than the enforcement actions and funding-flexibility bills of recent weeks. It does not regulate a vendor or re-permission a funding stream. It assigns institutional ownership.

The structural implication for public education is that the locus of AI governance is moving from the document to the org chart. A board-ratified AI policy with no named owner has been the national default. Maryland has now made the named owner a matter of law, and the design of that role, where it sits, what authority it carries, and what it is accountable for, becomes the governance question that matters. Districts in the other forty-nine states should read Maryland not as a local development but as a template. When a state with Maryland's policy influence converts an AI coordinator from a best practice into a mandate, vendors, counsel, and neighboring legislatures begin to treat the role as the expected standard of care, whether or not their own statutes yet require it.

The same week supplied the counter-pressure that makes this urgent. In New Mexico, a Legislative Education Study Committee analyst told lawmakers on May 28 that the state's 2025 AI guidance is not binding and recommended a formal oversight body to vet tools and publish model policies. And the peer-reviewed research that surfaced in this window points to what these governance roles will actually have to manage. A cross-sectional study of senior high school students found that the awareness and ethics dimensions of AI literacy predicted academic performance more strongly than the technical skills dimension, while most students' AI learning occurred informally, outside any teacher-guided structure. Read together, the signals describe a single problem. States are now naming the people who must govern AI in schools faster than anyone has produced evidence about what those people should do. The institutional architecture is being mandated before the evidence base that should inform it exists, which means the districts that design the role well, rather than merely fill it, will be the ones that turn a compliance requirement into an actual governance capability.

Top Research and Policy Signals

1. Maryland Enacts the Artificial Intelligence Ready Schools Act, Mandating a District AI Coordinator in Every Local School System

Source type. State legislation (enacted; signed by the Governor, May 26, 2026, Chapter 634).

Maryland's SB 720 amends the state Education Article to require the Maryland State Department of Education (MSDE) to provide AI guidance to local school systems, educators, parents, and students through an online platform, and to develop guidelines, best practices, and evaluative tools to help districts assess the AI tools they select. The act requires each local school system to designate a non-instructional central-office coordinator to manage AI use, requires professional development for teachers and school leaders, and establishes the Maryland AI Education Collaborative, a multi-stakeholder body of educators, administrators, parents, students, and organizational representatives charged with studying AI use and recommending guidance, professional development, and policy. The bill passed with near-unanimous margins (Senate 45-0 on final passage, House 129-8) and directs that AI literacy be integrated into the state's workforce readiness and computer science standards by June 1, 2027. The bill carries a sunset provision, signaling the legislature intends to revisit the framework as evidence accumulates.

Leadership implication. Maryland has made the question that every cabinet has been avoiding, who owns AI governance, into a legal requirement. Superintendents in any state should not wait for their own legislature to act before answering it. Name the office now, define its authority over procurement and policy in writing, and give it standing to evaluate tools against criteria the district controls. The districts that treat the coordinator role as a real governance seat, rather than a title added to an existing job description, will be the ones positioned to absorb the mandate when it reaches their state.

2. New Mexico Legislative Working Group Recommends a Formal State AI Oversight Body, Citing Non-Binding Guidance

Source type. State legislative committee presentation and contemporaneous reporting (May 28-29, 2026).

At a Legislative Education Study Committee meeting, a policy analyst presented a report recommending that New Mexico formally assign responsibility for AI guidance to a multidisciplinary oversight body, a new or existing council, board, independent office, or Public Education Department bureau, tasked with vetting AI tools, publishing model AI policies, and providing AI-literacy professional development. The recommendation rests on a specific structural problem: the state issued AI guidance in 2025, but schools and districts are not required to follow it. The report cited concerns about student data privacy, the use of AI for cheating, and an overreliance that diminishes cognitive engagement. The recommendation is not yet law; it is a working-group proposal to lawmakers ahead of the next legislative session.

Leadership implication. The distance between state guidance and binding state policy is exactly the gap where district-level exposure accumulates. New Mexico's own analyst is now saying out loud that non-binding guidance does not produce governance. District leaders should not interpret the existence of state guidance as cover. Until a binding framework exists, the operative AI governance authority in most states is the local board, and the absence of a district policy is not neutral; it is an unmanaged risk.

3. Peer-Reviewed Study: Awareness and Ethics, Not Technical Skill, Are the Strongest AI-Literacy Predictors of Academic Performance

Source type. Peer-reviewed, published in the International Journal of Technology in Education (2026). Cross-sectional correlational study of 525 senior high school students across four academic strands. Identified via Consensus.

Using the validated Artificial Intelligence Literacy Scale, the authors measured AI literacy across four dimensions (awareness, usage, evaluation, and ethics) and correlated it with grade point averages and standardized test scores among 525 senior high school students. They found significant positive relationships between AI literacy and academic performance, with the awareness and ethics dimensions emerging as the primary predictors, ahead of technical usage skill. AI literacy explained roughly 7 percent of the variance in both grades and test scores. Despite 85.7 percent of students reporting that they use ChatGPT, the authors characterize students' AI learning as predominantly informal and peer-driven rather than teacher-guided, and find performance variation across academic strands. As a cross-sectional correlational design, the study identifies associations rather than causal effects, and its sample is drawn from a single national context rather than the U.S.

Leadership implication. This is direct evidence against the most common shape of school AI-literacy programming, which front-loads tool mechanics. The dimensions that track with performance are conceptual: knowing what AI is, where it shows up, and how to evaluate and use it ethically. District AI-literacy scope and sequence should lead with awareness and ethics, not keystrokes. And the finding that most student AI learning is informal and peer-driven is a governance flag in its own right: if the district is not teaching this, students are still learning it, just without guidance.

4. Peer-Reviewed Study: High School Students Split Between Critical Engagement and Passive Dependence When Using Generative AI in English Language Arts

Source type. Peer-reviewed, published in Computers and Education: Artificial Intelligence (2026). Mixed-methods case study of high school students in an English Language Arts course. Identified via Consensus.

This mixed-methods case study examined how high school students interacted with an AI-enhanced resume-writing unit and a text-based mock-interview unit inside an English Language Arts course. Students reported positive perceptions of AI utility across awareness, usage, evaluation, and ethics, and many engaged critically, questioning, personalizing, and revising AI suggestions. But a meaningful subset adopted a passive or overly dependent approach. Students also identified concrete limitations: generic or decontextualized AI responses, difficulty interpreting AI feedback, and ethical concerns about authenticity, plagiarism, and overreliance. The authors conclude that meaningful integration requires explicit instruction in prompting, verification, and critical evaluation so that AI functions as a scaffold rather than a substitute for learning. As a single-course case study, the findings are rich but not generalizable to all classrooms or subjects.

Leadership implication. The same assignment produced two different student behaviors, critical use and passive dependence, which means the variable that matters is not the tool but the instructional design around it. This is non-STEM, ELA-specific evidence that districts can use to push AI literacy out of the computer-science silo. Procurement and curriculum decisions should require that AI-enabled assignments include explicit instruction in prompting, verification, and evaluation, because without them, the same tool that scaffolds one student's learning will substitute for another's.

5. Preprint: A Large Randomized Trial Finds Prompting-Literacy Instruction Improves Students' Ability to Use AI as a Tutor Rather Than an Answer Engine

Source type. Preprint (not peer-reviewed), posted to arXiv (2026). Semester-long randomized controlled trial, N = 979, in a college-level introductory computer science (CS1) course. Higher-education context; transfer to K-12 is inferential. Identified via Consensus.

This semester-long randomized controlled trial assigned 979 students across four instructional conditions of increasing cognitive engagement, grounded in the ICAP framework, to teach prompting literacy: using AI as a tutor rather than as a solution provider. All conditions significantly improved prompting skill, with gains rising progressively as engagement intensity increased, validating the framework's hierarchy. Among students with similar pre-test scores, higher immediate learning gains predicted higher final-exam scores, though no direct between-group differences in exam performance emerged. The authors frame the work as a scalable way to convert AI-use policy into actual instruction. Two limits matter for K-12 leaders: this is a preprint that has not completed peer review, and the population is college CS1 students, not K-12 learners, so the transfer to younger students is a hypothesis, not a finding.

Leadership implication. The actionable signal here is structural, not specific: prompting literacy can be taught at scale, and how it is taught, the degree of cognitive engagement, changes how much students gain. For districts, the lesson is that an AI-use policy is not instruction. A policy that says use AI as a tutor, not an answer key, accomplishes nothing unless it is paired with a taught, scaffolded module. Treat this as preprint evidence to watch, not yet as a basis for procurement, and require any vendor claiming to build prompting literacy to show how, and at what level of engagement, students actually practice it.

Emerging Strategic Themes

Theme 1. Governance Is Moving From the Document to the Org Chart. Maryland's mandated district AI coordinator marks a shift in what state AI policy regulates. Earlier waves regulated tools, funding, and disclosure. This one assigns named institutional ownership. The strategic question for districts is no longer only what the policy says, but who is accountable for it, what authority that person holds over procurement and curriculum, and whether the role is resourced to function.

Theme 2. Non-Binding Guidance Is Being Recognized as Insufficient by the States That Issued It. New Mexico's own legislative analyst now argues that the 2025 guidance did not establish governance because districts are not required to follow it. Expect more states to move from guidance toward binding frameworks and oversight bodies. Districts that built real policy on top of state guidance will absorb that shift easily; districts that treated guidance as a substitute for policy will be exposed when the binding version arrives.

Theme 3. The Evidence Says Lead AI Literacy With Concepts, Not Tools. Two independent peer-reviewed findings this week, that awareness and ethics predict performance more than technical skill, and that the same AI assignment produces both critical use and passive dependence, point in the same direction. The differentiator is conceptual and instructional design, not access. AI-literacy programs anchored in tool mechanics are optimizing the wrong variable.

Theme 4. Policy Without Instruction Is a Null Action. The prompting-literacy trial, even as a preprint, reinforces a theme running through all of this week's research: stating a rule about how students should use AI changes nothing unless the rule is taught. "Use AI as a tutor, not a crutch" is a slogan until it becomes a scaffolded, practiced module. Districts should audit whether actual instruction backs their AI-use policies or whether they are aspirational text.

What Was Not Found

Four evidence categories did not appear in this week's window, and each absence carries a present-tense cost while districts and states are acting.

First, no study, peer-reviewed or preprint, has tested whether designating a district AI coordinator, the precise mechanism Maryland just made mandatory, improves any student outcome, equity measure, or safety result. States are now legislating a governance role with no evidence of outcomes behind it. Districts mandated to fill the role are doing so without empirical guidance on how to design it for effect rather than compliance.

Second, no peer-reviewed U.S. population study this week measured AI literacy outcomes for the subgroups that governance frameworks claim to protect: English learners, students with Individualized Education Programs, and students in high-poverty districts. The strongest AI-literacy evidence in this window, the 525-student cross-sectional study, is drawn from a single non-U.S. national context, is correlational rather than causal, and does not stratify by these populations. Districts are developing an AI-literacy scope and sequence for these students by drawing on associations found elsewhere.

Third, no causal study isolated the effect of the conceptual dimensions of AI literacy, awareness and ethics, on learning. The cross-sectional finding that these dimensions predict performance is an association that could run in either direction; stronger students may develop stronger AI awareness. Districts redesigning curriculum around awareness and ethics should know they are acting on a correlation, not a demonstrated cause.

Fourth, no peer-reviewed K-12 evidence confirmed that prompting-literacy instruction transfers to younger students. The strongest prompting-literacy result this week is a preprint conducted with college students. The mechanism is plausible for K-12, but plausibility is not evidence, and districts purchasing prompting-literacy products for middle and high school are buying ahead of any age-appropriate validation.

Novo Executive Summary

This was the week the question of AI governance ownership stopped being optional. Maryland made the district AI coordinator a matter of law; New Mexico's legislature heard that non-binding guidance is not governance; and peer-reviewed research made clear that what these governance roles must manage is conceptual and instructional, not technical. The throughline is that institutional architecture, named ownership, defined authority, and taught instruction, is now the governing variable, and states are mandating it faster than the evidence base can mature. A policy without an owner is a document; a rule without instruction is a slogan; guidance without enforcement is a suggestion. Districts that close those three gaps before their state forces the issue will hold a real advantage when it does. Novo Innovative Pathways works with district cabinets to design exactly this architecture, the governance ownership model, the role-based AI literacy that leads with concepts and ethics, and the implementation strategy that turns policy language into taught practice, so that when the mandate arrives, the capability is already in place.

Watch This Week

• New York City Department of Education comprehensive AI playbook, expected to be released in June 2026 following its public feedback period, including a tool-vetting process screening for algorithmic bias, equity impact, and developmental appropriateness.
• Maryland State Department of Education development of the SB 720 guidance platform, evaluative tools, and the standup of the Maryland AI Education Collaborative.
• Ohio statutory deadline for district AI policy adoption, July 1, 2026.
• New Mexico Legislative Education Study Committee consideration of the recommended AI oversight body ahead of the 2027 legislative session.
• Continued movement of 2026 state AI-in-education bills tracked by FutureEd and the PIE Network, including comprehensive frameworks in Oklahoma and parental-consent proposals in South Carolina.

Sources

Governance and Policy

Research, Peer-Reviewed

Research, Preprint (Not Peer-Reviewed)