Florida Moves to Govern Classroom AI Through the Internet Safety Rule: Consent Becomes the Unit of K-12 AI Governance

Framing

The most consequential AI governance move of the week did not arrive as an AI law. It arrived as an amendment to a rule most district leaders file under routine compliance. On July 21, the Florida Department of Education noticed a proposed change to Rule 6A-1.0957, the internet safety policy rule that every district school board and charter governing board in the state already reviews and re-adopts each year. Under the draft, by January 1, 2027, every one of those boards must fold AI governance into that policy: plain-language parental notice for every approved AI instructional tool, opt-in consent with a non-AI alternative before a student uses one directly, an option for parents to limit usage time, a publicly accessible list of approved tools, teacher and administrator training, interaction records that satisfy federal parental access rights, and annual reporting to the state of every tool in use, including how often and how long students use it. The definition section quietly does its own governing: a tool designed to simulate friendship or use relationship-building engagement features cannot qualify as an AI instructional tool at all.

Read the mechanism, not just the content. Florida is not building new AI governance machinery. It is loading AI into machinery that already exists, a CIPA-era compliance instrument with an annual adoption cycle, an existing enforcement posture, and a two-decade habit of board sign-off. That is faster than legislation, cheaper than a model policy campaign, and harder for a district to ignore, because nobody skips the policy their E-Rate funding history taught them to file. This brief has tracked statutes, agency guidance, board policy, and, last edition, collective bargaining as governance layers. This week adds a quieter fifth channel: the compliance documents districts already own. Whoever writes AI into those documents first, the state or the district, sets the defaults everyone else inherits.

The research this week says most districts are losing that race inside their own paperwork. A nationally representative study in Educational Policy finds nearly two-thirds of student handbooks never mention AI, and a second new study in the same journal finds the twelve largest districts actively rewriting integrity definitions, setting standards for users and vendors, and running deliberate experiments. The distance between those two findings is the governance gap, and it is demographically patterned: the districts least likely to have AI language are the more racially diverse and economically disadvantaged ones. Meanwhile, an analysis of 11,020 students accepted in Frontiers in Psychology shows student AI literacy stratifying by family resources, parental involvement, and school support before any policy is written. The Florida fight, at bottom, is about who holds the pen when the defaults get set: the state, the vendor coalition that filed its objections on July 29, or the parents the opt-in provision would newly empower. Districts that have not inventoried which of their own documents govern AI are not at that table. They are on it.

Top Research and Policy Signals

1. Florida Puts AI Inside the Internet Safety Policy, and Sets Consent at Opt-In

Source type. State regulatory action. Notice of Development of Rulemaking published July 21, 2026; rule development workshop August 5, 2026.

Florida Department of Education. (2026, July 21). Notice of development of rulemaking: Rule 6A-1.0957, Internet Safety Policy. Florida Administrative Register, 52(140).

The draft amendment requires every district school board and charter governing board to adopt an AI amendment to its internet safety policy by January 1, 2027. Thirteen minimum requirements follow. The ones that will reorganize district operations: parents must receive plain-language notice naming each approved AI instructional tool, its grade levels and subjects, and the nature of student interaction; parents must opt their child in to direct use of a tool, with a non-AI alternative provided for students not opted in, and may limit usage time; districts must keep a public, regularly updated list of approved tools; AI tools must maintain records of student interactions to support parental access rights under FERPA; districts must report every tool, with usage frequency and interaction duration, to the state annually; approved tools may not sell, monetize, profile, or use student data to train commercial AI models, and vendors that keep data in the United States get priority. The definitions bar tools with companionship or anthropomorphic engagement design from qualifying as instructional tools at all.

On July 29, SIIA, the ed-tech industry's principal trade association, filed comments calling the AI definition overbroad and warning that opt-in plus alternative assignments would create significant administrative burden while disproportionately cutting off struggling readers, English language learners, and students with disabilities from adaptive tools.

Leadership implication. Even outside Florida, this is the template to study, because regulating through an existing compliance instrument travels fast between states and requires no legislative session. Price the operational load now: consent tracking, alternative assignments, interaction recordkeeping, and usage-duration reporting are staff work, not policy words. If your vendor contracts cannot produce per-student interaction records or usage-duration data on request, please renegotiate before the rule takes effect, not after.

2. Two Thirds of Student Handbooks Say Nothing About AI

Source type. Peer-reviewed. Educational Policy, published online June 25, 2026.

Curran, F. C., Goo, J., Thomas, C., Wang, P., Moylan, K., & Maharaj, S. (2026). Collaborator or accomplice? A national examination of artificial intelligence (AI) in school district discipline policy. Educational Policy. Advance online publication.

Using a nationally representative sample of school district student handbooks and codes of conduct, the study finds that nearly two-thirds of districts do not reference AI at all. Students in those districts operate without written guidance on what AI use is permitted, what constitutes misuse, and what discipline may follow. References to AI vary across district demographics, with more racially diverse and economically disadvantaged districts less likely to reference AI in their handbooks. The authors document exemplar handbook integrations and lay out the legal and policy case for treating AI as a discipline-policy matter rather than an instructional footnote.

Study context. National sample of publicly available district handbooks and codes of conduct, University of Florida research team, advance online publication.

Leadership implication. When the handbook is silent, AI discipline decisions get made case by case, teacher by teacher, and that is where inconsistency, appeals, and disparate-impact exposure live. The demographic pattern converts a paperwork gap into an equity finding: the students statistically most exposed to discipline are concentrated in districts least likely to have written the rules down. Your student handbook is a governance document. This study says most districts have not treated it like one.

3. How the Twelve Largest Districts Are Actually Governing

Source type. Peer-reviewed. Educational Policy, advance online publication, 2026.

Liang, S., et al. (2026). Governing GenAI through redefinition, regulation, and innovation: Policy responses in the largest US school districts. Educational Policy. Advance online publication. [Full author list and article URL flagged for verification; abstract confirmed via the Consensus academic index]

Analyzing policy documents from the twelve largest US school districts, the study finds systems that have moved from reactive bans to proactive management through three interrelated moves: redefining foundational concepts such as academic integrity, regulating through standards written for both users and vendors, and innovating through intentional experimentation. The authors offer a typology of district-level AI governance and position it as a baseline for the causal and implementation research that does not yet exist.

Study context. Document and policy analysis of the twelve largest US districts by enrollment. Findings describe what large systems are doing, not whether it works.

Leadership implication. The typology is a usable self-audit. Ask which of the three moves your district has made in writing: have you redefined integrity for an AI era, have you set standards that bind vendors and not just students, and do you have a sanctioned lane for experimentation? Most districts have improvised pieces of one. The largest systems are doing all three deliberately, and their vendor-facing standards will shape the products that show up in your procurement pipeline either way.

4. AI Literacy Is Stratifying Before Policy Arrives

Source type. Peer-reviewed. Frontiers in Psychology, accepted May 11, 2026; advance abstract online, final formatted version pending.

Wang, D., Wang, Q., & Zhang, S. (2026). Latent profiles of AI literacy among K-12 students: Predictors and links to self-regulated learning. Frontiers in Psychology, 17, Article 1834851.

Drawing on survey data from 11,020 K-12 students in China, the study uses latent profile analysis to identify four distinct, progressively ordered AI literacy profiles, from foundational-limited to high-excellence. Profile membership is not random: gender, only-child status, educational stage, family socioeconomic status, frequency of AI use, parental active mediation, and school AI support all significantly predict where a student lands. Students in higher literacy profiles show consistently stronger self-regulated learning.

Study context. Large sample, single country, self-report survey, cross-sectional design. The stratification pattern is the finding; causal direction is not established.

Leadership implication. Literacy is not a curriculum box to check; it is a distribution your district inherits on day one, shaped by family resources and school support. Two levers in this study belong to you: school AI support and the parent-mediation channel, which most districts have never treated as an instructional partner. If your AI literacy plan assumes a uniform starting line, this study says it will widen the gap it was funded to close.

5. The Students at the Center of the Consent Fight

Source type. Peer-reviewed conceptual review. Learning Disability Quarterly, published online April 19, 2026.

Seung, Y., & Basham, J. D. (2026). Cognitive offloading in the age of generative AI: What does it mean for students with learning disabilities? Learning Disability Quarterly. Advance online publication.

This conceptual review examines generative AI for students with learning disabilities through the lens of cognitive offloading, the delegation of cognitive tasks to external tools. The tension it maps is the one that matters for policy: the same tool that functions as a legitimate compensatory aid, the way a calculator or text-to-speech does, can also bypass the cognitive processes a student most needs to develop when reliance becomes excessive. The authors synthesize what shapes students' offloading decisions and propose an initial conceptual model for strategic, rather than wholesale, offloading. What the field does not have, and the authors are direct about this, is empirical evidence on how students with learning disabilities actually make these decisions and what the learning consequences are.

Study context. Conceptual review and model, not an outcome study. It defines the research agenda; it cannot yet settle the practice questions.

Leadership implication. Notice who sits at the center of the Florida collision: SIIA argues opt-in consent will cut students with disabilities off from adaptive tools. At the same time, the offloading literature warns that unstructured access carries its own risk for exactly these students. Both can be true, which is why the decision belongs inside the IEP process, individualized and documented, not inside a blanket district default. If your special education team has no written stance on AI as an accommodation, both the consent forms and the grievances will arrive before the evidence does.

Emerging Strategic Themes

Theme 1. Governance by existing instrument. Florida's move shows states can regulate classroom AI through compliance machinery districts already operate: internet safety policies, handbooks, codes of conduct, materials-review processes. No new statute, no new agency, an annual adoption cycle already in place. Audit every document your district re-adopts each year and assume each one is a candidate AI governance vehicle, because to a state rule-writer, that is exactly what it is.

Theme 2. Consent is becoming the unit of governance. Florida's draft sets opt-in. Oklahoma's new statute sets opt-out with annual disclosure of tools and data collected. Both convert AI use from an instructional decision into a family-by-family authorization workflow, with alternative assignments as the fallback. Districts should design the consent infrastructure, forms, tracking, and the non-AI alternative pathway now, because the default may be decided for you and the administration of it will not be.

Theme 3. The document gap is an equity gap. Two-thirds of handbooks are silent on AI, and the silence concentrates in racially diverse and economically disadvantaged districts, echoing last month's finding that policy quality tracks district wealth. Layer the literacy stratification data on top, and the pattern is consistent: the students with the least structured support are governed by the thinnest documents. Written governance is becoming a resource that is unequally distributed, and it will be read that way in public.

Theme 4. The accommodation dilemma moves to center stage. Vendors invoke students with disabilities to argue against restrictions; the offloading literature invokes the same students to argue for structure. The honest position is that AI-as-accommodation currently runs on professional judgment, not outcome evidence. Districts that route these decisions through IEP and 504 teams with explicit documentation will be defensible; districts that let the default decide will eventually explain themselves to a hearing officer.

What Was Not Found

This section reports what the evidence base still cannot support, because decisions are being made this week that assume otherwise.

• No study exists on consent defaults for educational AI. Nothing measures how opt-in versus opt-out affects participation rates, which families opt out or fail to opt in, or what happens to the learning of students routed to alternative assignments. Florida and Oklahoma have now chosen opposite defaults for the same decision. Neither can point to evidence, because there is none.
• No evidence connects internet-safety-style controls to AI outcomes. Filters, notices, and time limits have two decades of implementation history for web content, but no study tests whether any of them changes AI-related risks or learning when applied to AI tools. Florida is extending a familiar instrument to an unfamiliar technology on structural analogy alone.
• No study links handbook or policy language to student behavior. The Curran team documents the silence and its demographic pattern. What no one has measured is whether AI language in a handbook changes what students do, how consistently teachers respond, or who gets disciplined. Districts also do not publish AI-related discipline counts by race or disability status, so no one can currently check the disparate-impact question this study raises.
• No United States measurement of student AI literacy at scale. The strongest stratification evidence this week comes from 11,020 students in China, self-reported and cross-sectional. There is no nationally representative United States dataset measuring student AI literacy at all. States are mandating AI literacy instruction for a construct no one is measuring at population scale.
• No outcome study of generative AI for students with disabilities. The accommodation debate now shaping consent rules runs entirely on conceptual models and professional judgment. For the second consecutive edition, this window produced no empirical K-12 outcome data for the population most invoked on both sides of the argument.
• No cost or capacity data for the new compliance load. Nothing estimates what consent tracking, interaction recordkeeping, alternative-assignment staffing, or usage-duration reporting costs a district to run. Florida's January 1, 2027 deadline assumes an administrative capacity that no one has priced, least of all for small and rural systems.

The pattern is the same one this brief has documented for months, but this week it sharpened: states are now writing operational mandates, with dates and reporting requirements, on questions where the research base cannot yet answer whether the mandated practice helps, harms, or merely costs. That is not a reason to wait. It is the reason districts need monitoring, documentation, and exit ramps built into everything they adopt, because the evidence will arrive after the obligations do.

Novo Executive Summary

Florida just demonstrated that the fastest route to binding AI governance runs through documents districts already maintain, and every state rule-writer watching now knows it. The operational lesson for district leaders is to stop thinking of AI governance as a new document to write and start treating it as an inventory to control: the internet safety policy, the student handbook, the code of conduct, the vendor contract, the IEP. Each is a surface where AI defaults are about to be set, by you or for you. This week's research shows most districts have not claimed those surfaces, that the silence is deepest in the districts with the least slack, and that student readiness is stratifying while the paperwork waits. Consent workflows, interaction records, and usage reporting are coming as compliance obligations ahead of any outcome evidence, which means architecture, not enthusiasm, is what separates districts that absorb the mandates from districts that they absorb. Novo Innovative Pathways works with district leaders on exactly this: governance architecture, role-based AI literacy, and implementation strategy, built into the documents you already own before someone else writes them for you.

Watch This Week

• Wednesday, August 5: Florida Department of Education rule development workshop on the 6A-1.0957 amendment. Watch whether the AI definition narrows toward general-purpose chatbots and whether opt-in survives vendor and parent-group pressure from opposite directions.
• Monday, August 3 and Wednesday, August 5: California Senate and Assembly appropriations committees hold rapid-fire hearings deciding the fate of roughly 30 surviving AI bills, including AB 1159, which would extend student privacy law to school-marketed digital operators, AB 2071 on digital health curriculum, and AB 2656, which would require 45 days' notice to employee organizations before public employers deploy generative AI.
• Florida's next procedural step after the workshop: a formal Notice of Proposed Rule and public comment window, the stage where definition and consent language get locked.
• Ohio districts open the 2026-27 school year this month as the first full cohort operating under state-mandated district AI policies. Early implementation reports will show whether adopted policies function or sit in binders.
• Maryland's 120-day district policy clock under the AI Ready Schools Act continues running toward fall adoption deadlines across its 24 districts.
• New York City's final AI playbook, promised by September, is still pending, with the citywide educational software purchasing pause still standing until it lands.

Sources

Governance and Policy

Research, Peer-Reviewed

Research, Preprint (Not Peer-Reviewed)

No preprint met the inclusion bar in this week's window. Preprint candidates surfaced in the search were either previously covered in earlier editions or fell outside the K-12 governance scope of this issue.

AI in Public Education Brief is published weekly by Novo Innovative Pathways. For district advisory engagements, contact Dr. Reginald Griffin through Novo Innovative Pathways.