Skip to main content
OpenEduCat logo
Updated Quarterly · Q3 2026

AI Education Policy Tracker

UNESCO frameworks, OECD recommendations, the EU AI Act’s education provisions, US federal guidance, and a fast-moving map of state law — often pulling in different directions. This tracker translates each instrument into three institutional columns: what you must do, what you should do, and what to watch.

Vendor-neutral editorial reference. Links go to primary sources; verify before acting. Not legal advice.

GlobalAdopted framework guidance across UNESCO member states

UNESCO — Guidance for Generative AI in Education and Research (2023, with subsequent updates)

The reference framework most national policies cite. Its core requirements: mandatory human oversight of AI-generated content in educational contexts, age-appropriateness safeguards, teacher capacity-building before student-facing deployment, and a prohibition stance on fully automated high-stakes assessment decisions.

Must do

If your national ministry references UNESCO guidance (most do), document human-review steps for any AI output that affects students — and be able to show them.

Should do

Map your AI tools against the guidance’s teacher-training expectations; UNESCO treats staff AI literacy as a precondition, not an afterthought.

Watch

UNESCO’s AI competency frameworks for students and teachers are becoming the default curriculum reference — expect national curricula to absorb them.

Primary source: unesco.org

OECD countriesRecommendations; heavily referenced by ministries of education

OECD — AI Principles and education-sector digital policy work

The OECD’s AI Principles (updated 2024) anchor public-sector AI expectations: human-in-the-loop oversight, accountability structures, and transparency. Its education policy work tracks AI governance as a top stated priority across member education ministries, with a growing number running national AI-in-education policy processes.

Must do

Nothing directly binding — but if you are in an OECD member state, your ministry’s guidance almost certainly inherits these principles. Align procurement criteria with them now.

Should do

Adopt the OECD’s human-oversight and accountability vocabulary in your institutional AI policy; it makes ministry and accreditor conversations easier.

Watch

OECD comparative reporting on national AI-education policies — useful early warning for what your ministry will likely require next.

Primary source: oecd.ai

European UnionIn force; high-risk obligations phasing in through 2026-2027

EU AI Act (Regulation 2024/1689) — education provisions

The AI Act classifies AI systems used for admission, assessment, progression, and proctoring-type monitoring in education as high-risk (Annex III). High-risk systems require conformity assessment, human oversight, logging, and transparency. Advisory systems where humans review each consequential action are designed to operate below the autonomy threshold that triggers the heaviest obligations — but classification depends on actual use.

Must do

EU institutions: inventory every AI system touching admission, assessment, or progression decisions, and determine with counsel whether each falls in Annex III. Deployers of high-risk systems carry their own obligations (human oversight, monitoring, record-keeping).

Should do

Prefer advisory-mode architectures with logging for anything near the high-risk boundary — the compliance distance between "AI decides" and "AI drafts, staff decide" is enormous.

Watch

Guidance and standards specifying how conformity assessment applies to education deployments, and member-state enforcement postures as obligations phase in.

Primary source: artificialintelligenceact.eu

United States — federalNon-binding federal guidance; FERPA/COPPA obligations remain binding

US Department of Education — AI reports and guidance (2023 onward)

The Department’s "Artificial Intelligence and the Future of Teaching and Learning" report set the federal tone: humans in the loop for consequential decisions, transparency disclosures, and vendor accountability. FERPA guidance points toward audit trails demonstrating staff review when AI acts on education records. No comprehensive federal AI-education statute exists; existing privacy law does the binding work.

Must do

Apply FERPA/COPPA analysis to every AI tool that touches student data — including free teacher tools adopted classroom-by-classroom. Require audit capability from vendors.

Should do

Use the federal report’s human-in-the-loop framing as the backbone of your district or campus AI policy; it is the de facto national reference.

Watch

Federal procurement and civil-rights guidance touching algorithmic decisions in education — the enforcement lever most likely to move first.

Primary source: ed.gov

United States — statesFast-moving: a majority of states now have enacted or introduced measures

State AI-in-education legislation and executive guidance

State activity has accelerated sharply — from a handful of states with AI-education guidance in 2024 to a substantial majority with legislation, task forces, or department-of-education guidance today. Common threads: required district AI policies, teacher-training expectations, vendor transparency requirements, and restrictions on automated consequential decisions.

Must do

Check your own state department of education’s AI guidance — several states now expect districts to maintain a written AI policy with named components.

Should do

Draft your policy against the common-thread checklist (acceptable use, data governance, training, human review, audit cadence) so state-specific requirements become edits rather than rewrites.

Watch

Algorithmic-accountability bills that reach beyond schools but capture education vendors — their transparency clauses often apply to AI used on minors.

Primary source: Your state DOE + CoSN policy tracking

The common thread: human oversight, logged

Across every instrument on this page, the same two requirements recur: a human must review consequential AI actions, and the institution must be able to demonstrate it. That convergence is worth designing for. Platforms whose AI layer is advisory by architecture — with guardrails bounding scope, budgets bounding cost, and audit logs recording every interaction — satisfy the oversight pattern these policies impose. OpenEduCat’s AI layer is built on exactly that pattern; see how advisory agents work and the AI-powered LMS for the implementation.

Frequently Asked Questions

Common questions from compliance officers, academic leadership, and boards tracking AI policy.

UNESCO’s Guidance for Generative AI in Education and Research (2023) is the global reference: it calls for human oversight of AI-generated content in education, protections appropriate to learners’ ages, teacher capacity-building before deployment, and it takes a prohibitive stance on fully automated high-stakes assessment decisions. Its companion AI competency frameworks for students and teachers are increasingly absorbed into national curricula. Verify current versions on unesco.org, as the guidance is periodically updated.

Building your institutional AI policy?

Start from the policy-component checklist in our glossary, then see how named guardrails and audit logs make each clause enforceable.

AI Policy ChecklistBook a Demo