AI Policy Frameworks Every School District Should Have

Why governance comes before tools

The districts navigating AI well are not the ones with the biggest tech budgets — they are the ones with clear, board-approved governance that lets them say yes or no to a new tool in days, not months. Without that scaffolding, every new AI request becomes a one-off debate and the answer defaults to "not yet," which pushes usage into the shadows.

This post covers the four documents we recommend every district have in place before the next school year, with a starter outline and the most common failure modes for each.

1. AI Acceptable Use Policy (AUP)

Audience: students, teachers, staff, contractors.

Must answer:

• Approved-tool list and the process for requesting additions.
• Required disclosure language when AI is used in instruction or assessment.
• Prohibited uses (e.g., pasting student PII, generating disciplinary letters without human review).
• Consequences for misuse, graduated by intent and harm.
• Annual review and acknowledgment cadence.

Common failure mode: writing it as a ban list. Ban lists age out in a quarter. Write it as a decision framework so it stays useful.

2. AI Procurement Standard

Audience: anyone who can sign a contract or click "install."

Must answer:

• Minimum SafeGradeAI tier required for student-facing tools.
• Required DPA clauses: training-data exclusion, retention window, sub-processor disclosure, breach notification SLA.
• Pedagogical review: does this tool advance an articulated learning outcome?
• Equity review: cost to families, accessibility, language coverage.
• Sunset clause: tools must be re-reviewed every 12 months or on major change.

Common failure mode: legal signs the DPA but no one re-reviews when the vendor changes their privacy policy. Build in the recurring check.

3. Data Handling Addendum

Audience: every staff member who touches student data.

Must answer:

• Which data classes can be entered into AI tools (almost always: none containing PII or FERPA-protected records unless the tool is under a school-official DPA).
• How AI-generated artifacts (lesson plans, feedback drafts, IEP suggestions) are stored, labeled, and retained.
• Who owns student-generated outputs and how long they're kept.
• Incident response: what to do if PII is pasted into a non-approved tool.

Common failure mode: assuming staff intuitively know what counts as PII. Provide a one-page examples sheet with "do paste / do not paste" rows.

4. Classroom Disclosure Guide

Audience: teachers, with sample language for families and students.

Must answer:

• When teachers must disclose AI use to families (e.g., any tool that processes student work).
• Sample family letters in plain language and at multiple reading levels.
• Student-facing disclosure: "This activity uses an AI assistant. Here is what it sees, here is what it does not."
• Opt-out pathway and the equivalent non-AI alternative.

Common failure mode: no equivalent non-AI alternative, which makes the opt-out theoretical.

A 30-day rollout sequence

Week 1: convene a working group (admin, teacher, IT, parent, student rep). Inventory current AI usage.

Week 2: draft AUP and Procurement Standard using the structures above. Pull from SafeGradeAI templates if you're a School Pro member.

Week 3: legal review. Family advisory review. Union review where applicable.

Week 4: board adoption, staff training, parent communication, and publish the approved-tool list.

What good looks like

• A teacher can check the approved list in 30 seconds.
• A new tool request gets a yes/no in 10 business days.
• Every approved tool has a named owner and a re-review date on the calendar.
• Families can read the AUP in under five minutes and know what to expect.

How SafeGradeAI School Pro helps

School Pro members receive board-ready templates for all four documents, the underlying rubric mapped to COPPA / FERPA / state-specific requirements, and monitoring alerts when an approved tool's score changes.