AI Policy & Guidance
Vague guidance is not policy.
Most district "AI policies" are a paragraph in the acceptable-use policy. Meanwhile AI is making or shaping decisions in HR screening, scheduling, early-warning systems, and instruction. If your policy only addresses ChatGPT in the classroom, it is governing a fraction of your actual AI use.
The Michigan picture
State guidance is a floor, not a ceiling
Michigan's MDE guidance (May 2026) is a meaningful first step: a phased implementation framework aligned to the MICIP continuous-improvement cycle, with the Statewide AI Workgroup's resources behind it. If your district hasn't read it, start there.
But notice what it doesn't yet give you: board-adoptable policy language, an enforceable prohibited-uses list, a risk-tiering rubric, a vendor review workflow, or an appeal process. Those gaps are not a criticism of the guidance — state guidance is designed to leave local decisions local. They are, however, exactly the pieces a district must build for itself. The resources on this page are the wraparound: they fill the gaps between what the state recommends and what a board can adopt and an administrator can enforce.
The framework
Twelve components of a complete district AI policy
From the AI Policy Playbook. If your current policy is missing more than a few of these, it has gaps a single incident will find.
Purpose & scope
What the policy governs, for whom, and why — including operational AI, not just classroom tools.
Definitions
Shared vocabulary so "AI," "generative AI," and "automated decision" mean the same thing in every building.
Human oversight
Where a human must review, decide, and own the outcome — non-negotiable for anything high-stakes.
Student data privacy
What data may touch which tools, under what agreements — the prompt box included.
Staff use
Approved tools, approved uses, and the professional-judgment rule.
Student use
Age-appropriate permissions, disclosure expectations, and instruction in responsible use.
Academic integrity & disclosure
Task-level clarity about what AI use is invited, permitted, and off-limits.
Equity, bias & accessibility
Access audits, bias review, and accessibility screens as policy, not aspiration.
Prohibited & high-risk uses
An enforceable list — with a risk-tiering rubric for everything in between.
Procurement & vendor review
Intake forms, vendor questionnaires, and data-processing agreements before purchase.
Transparency & public communication
What the community is told, in plain language, on a rhythm.
Review & continuous improvement
A standing cycle — policy that never gets revisited stops working.
Don't start blank
Learn from the strongest model policies
Four models worth borrowing from — each profiled in the Playbook with notes on its best use for Michigan districts.
Ohio's model policy
Legislatively mandated statewide model language — a strong starting text for board-adoptable wording.
New York City's traffic-light framework
Use-permission tiers students and staff can actually remember — green, yellow, red by use case.
North Carolina's EVERY framework
A principled implementation frame connecting AI guidance to instruction.
TeachAI's seven-principle toolkit
The most widely used national scaffold — principles that translate into local policy sections.
Sequence
First 30 / 90 / 365 days
The Playbook closes with a phased sequence for Michigan districts — workgroup and inventory in the first month, drafted policy and risk tiers by the first semester, board adoption and procurement integration within the year. The concern page walks through it.