A Practitioner Publication

The Educator's
AI Field Guide

What's safe, what's not, and how to decide

Version 1.0 · April 2026 · Nine sections

If you teach in Texas, the guidance you have been given about AI in your classroom has probably come from a vendor webinar, a compliance training slide deck, or a colleague down the hall. None of these is what you actually need. What you need is a field guide — written by a practitioner, for practitioners, that says what the rules are, names what the rules aren't, and gives you a decision you can make in ten seconds.

This playbook exists because a 1974 privacy law (FERPA), a 2026 AI statute (Texas HB 149), and your district's policy — if it has one — are each addressing a different part of the same problem, and none of them is writing for the teacher. The document below is.

Why I wrote it

I kept watching teachers hit the same three questions and not find clear answers.

After months of conversations with colleagues about AI tools, student data, and what counts as a safe account, I realized the answers I was giving in hallways and PLC meetings were the document that didn't exist yet. So I wrote it down, verified every vendor claim against primary sources, and built it to be the document I wish I had handed myself two years ago.

Table of Contents

Nine sections. Read in any order.

01

Why This Document Exists

The structural gap between three kinds of rules-makers
+

Between a 1974 privacy law and a 2026 AI statute sits a teacher holding a laptop at 4:45 on a Thursday. Nobody has written down, in plain language, what that teacher should do. Three forces — federal law (FERPA), state law (Texas HB 149), and district policy — are each addressing a different part of the problem, and none of them is writing for the classroom teacher.

1
Federal
FERPA · 1974 Written for filing cabinets. Sets the floor.
2
State
TRAIGA · 2026 Regulates commercial deployers, not classrooms.
3
District
Policy · Varies Often absent. Often out of date.

This section names the gap explicitly and positions the playbook as the document that fills it — written by a practitioner, for practitioners, in the space the other three forces leave empty.

02

The Three Questions Before Every AI Prompt

The operational core of the document
+

If a teacher tapes one page of this playbook above their desk, this is the one. Three questions, in the order to ask them. If any one of them gives a bad answer, the prompt doesn't go through.

1

Does this prompt contain any student PII — names, IDs, scores tied to a child, IEP content, metadata that could identify?

2

Is this account covered by a signed data agreement, or is it a consumer account that can't be made compliant with any toggle?

3

Is a human reviewing the output before it affects a student? AI drafts. The human decides.

The section includes a five-outcome flowchart that turns the three questions into a decision gate, and a test under each question a teacher can use in under five seconds.

03

FERPA, Translated for Teachers

A 1974 law, explained in working language
+

FERPA was passed in 1974, cited as 20 U.S.C. § 1232g and implemented at 34 CFR Part 99 — when a school principal kept grades in a locked filing cabinet and the word "upload" was not in the dictionary. None of that context has been updated. The law has not been meaningfully amended for the cloud era.

This section gives a teacher a working model of FERPA: a ten-item concrete list of what counts as an "education record," a then-vs-now view of how AI changed the calculus of disclosure, and an honest three-tier breakdown of what actually happens when something goes wrong — federal, district, and family. The core argument: FERPA is not a paperwork law, it is a trust law.

04

Account Types, Demystified

Consumer vs. institutional tiers, every major vendor
+

The same AI tool can be safe to use with student data on Monday and unsafe on Tuesday — depending entirely on which account you signed in with. This section is the reference table that turns that decision into a five-second check.

"ChatGPT is FERPA compliant"
"ChatGPT Edu is covered by our DPA"

The section includes a master matrix of ten account tiers across OpenAI, Google, Anthropic, and Microsoft — verified against primary vendor documentation in April 2026 — plus side-by-side vendor blocks detailing the exact distinction between each consumer and institutional tier.

05

Thirty Use Cases, Ranked

Green light, yellow light, red light — with specifics
+

The three questions, applied to real classroom moments. Some are safer than they feel; some are riskier than they look; all thirty are things a teacher somewhere is doing this week.

Green · 10 No student PII. Use any account. The output affects you, not a student record.
Yellow · 10 Student data involved. Covered account only. Human reviews every output.
Red · 10 Don't do this — regardless of account, prompt, or efficiency gain.

The red list opens with the highest-risk action a teacher can take: pasting an IEP, 504, or evaluation report into any AI. The yellow list closes with the clearest articulation of the drafting-vs-deciding distinction: AI drafts, you decide, every comment reads like it came from you because it did.

06

Three Districts, Three Playbooks

Same FERPA, very different Tuesdays
+

What a teacher can safely do with AI depends less on the tool and more on the district they teach in. This section gives three archetypes, a scene that makes each concrete, and a three-column playbook of what to do today, what not to do, and what to push the district toward next.

I Small rural Under 1,000 students · Part-time IT
II Mid-size suburban 5,000–15,000 · Evolving tools list
III Large urban 30,000+ · Formal vendor review

The rules are identical in all three. The paths to the tools that satisfy them are not. The section closes with five universal moves that work in an 800-student district and a 300,000-student district equally well.

07

Sample District AI Policy Framework

Ten articles. Two pages. Adopt, localize, review annually.
+

A district does not need a fifty-page AI policy. It needs a two-page one that is current, specific, and actually used. This is that one — adapt, approve, and attach an approved-tools list.

Art. IPurpose and Scope
Art. IIDefinitions
Art. IIIGeneral Principles
Art. IVEmployee Responsibilities
Art. VStudent Use
Art. VIProhibited Uses
Art. VIIApproved Tools List
Art. VIIIIncident Response
Art. IXTraining
Art. XReview and Revision

The framework includes fillable fields for district-specific details, instructional drafting notes in the margins, and a closing section explaining what was deliberately left out and why. Labeled as a "sample policy framework for district adaptation," not a model policy — the legal difference matters.

08

The 12-Question Self-Audit

Ten minutes. No one grades it but you.
+

Twelve questions, ten minutes, an honest picture of where your practice stands today. Print the page, circle yes or no, total the score, adjust the one or two things that stand out.

The question that matters most
"Have you, in the last 90 days, pasted any part of an IEP, 504 plan, or evaluation report into any AI tool?"

The scoring rubric reads from "confident practice" at 10–12 down to "stop and reset" at 0–3. The audit closes by naming question 6 — not question 12 — as the most telling answer. That one marks the line between thoughtful AI use and the kind that ends careers.

09

Resources & Citations

Every fact verified against a primary source
+

Every vendor claim in this playbook was verified against primary documentation in April 2026. This section organizes the citations by category — federal guidance, Texas state law, FERPA references, vendor docs, practitioner analysis — so readers can find what they need in seconds.

Federal US Dept of Ed · 2023 & 2025
State Law Texas TRAIGA · HB 149
FERPA 20 USC §1232g · 34 CFR 99
Vendors OpenAI · Google · Anthropic · Microsoft
Colophon

Prepared by Austin Lee, 6th Grade Advanced Math & Robotics Teacher and Instructional Leader, Cleburne ISD. All vendor-specific claims verified against primary documentation in April 2026.

This is a practitioner resource. It is not legal advice. It does not replace district policy. Where this document and a district's current policy disagree, follow the district.

Questions, corrections, or adaptation for your district?

This playbook was written to be forwarded, adapted, and improved. If your district is building AI policy and any of it is useful, use it. If something needs correcting, say so.

Get in touch