FERPA and the "School Official" Clause, Decoded

When OpenAI announced ChatGPT for Teachers in November 2025, the privacy language did something more specific than the usual "we take your data seriously." OpenAI designated itself a "School Official" with a "legitimate educational interest." Those aren't marketing words. They're the exact phrases the Family Educational Rights and Privacy Act uses to describe the one narrow path by which a school can hand student records to an outside company without asking every parent for permission first. Understanding that clause is the difference between trusting a vendor because it sounds trustworthy and knowing what it has actually bound itself to.

FERPA has governed student records since 1974. It's the reason a school can't fax your child's grades to a stranger, and the reason edtech companies spend so much effort proving they're not strangers. When OpenAI reaches for the School Official designation, it's stepping into a role with a legal shape — one that comes with duties, limits, and a very particular answer to the question every teacher should ask: if this goes wrong, who is in control of the data?

What "School Official" actually means

Under FERPA, a school generally can't disclose personally identifiable information from a student's education records without written consent. But the statute carves out exceptions, and the most important one for edtech is the school official exception. It lets a school share records — without consent — with a party that (1) performs a service the school would otherwise do itself, (2) is under the school's direct control over how it uses and keeps the data, and (3) uses the information only for the authorized purpose and doesn't re-disclose it.

The phrase "legitimate educational interest" is the qualifier that gates entry. A vendor doesn't get to rummage through records because it's curious; it gets access only to the extent needed to do the job the school hired it for. So when OpenAI names itself a School Official with a legitimate educational interest, it's not claiming a privilege — it's accepting a leash. The company is saying: treat us like a contractor you've brought inside the tent, bound by the same limits an employee would be.

The School Official clause doesn't make OpenAI a trusted outsider. It makes OpenAI a bound insider — one that only gets to touch student data on the school's terms, for the school's purposes, and under the school's control.

That framing matters because it flips the usual power dynamic. A generic privacy policy is something a company writes about itself. The School Official designation is something a statute writes about the relationship — and the school, not the vendor, holds the enforceable position.

Who actually owns the data

The most consequential line in OpenAI's education documentation is the plain one: student data "ultimately belongs to and remains under the control of the School." That single sentence is doing the heavy lifting the FERPA exception requires. Direct control by the school is not optional garnish — it's a condition of the exception being valid at all. Strip it out and the school official pathway collapses, and the disclosure would need consent it never collected.

Analysts tracking how FERPA maps onto 2026-era AI tools keep returning to the same test: can the institution decide what the vendor does with the data, and can it get the data back or deleted on its terms? If yes, the vendor is plausibly a school official. If the vendor reserves the right to use the data for its own ends, the relationship isn't a school-official relationship — it's a sale wearing a lanyard.

This is where OpenAI's no-training-by-default commitment stops being a nicety and becomes structurally necessary. If OpenAI trained its foundation models on whatever teachers typed into ChatGPT for Teachers, the data would be flowing to OpenAI's own purpose — model improvement — which is not the school's authorized purpose. Coverage of the launch's compliance posture, including reporting that frames it as FERPA-oriented and secure, treats the training carve-out as load-bearing for exactly this reason. The promise not to train isn't generosity. It's the price of staying inside the exception.

What the clause guarantees — and what it doesn't

Here's the honest ledger, because a designation is not a force field.

What it does buy you. A legal basis for the school to let teachers use the tool without collecting individual consent from every family. A vendor contractually and statutorily bound to use records only for authorized educational purposes. A no-re-disclosure obligation, so OpenAI can't pass student data down the chain. And a clear owner — the school — with the standing to demand deletion, audit use, or walk away.

What it doesn't buy you. The School Official clause says nothing about whether a teacher should paste a struggling student's full name and diagnosis into a chat window. The designation governs the vendor; it does not govern classroom judgment. FERPA compliance at the platform layer can coexist with a genuine privacy failure at the keyboard, if staff feed the system data it never needed. This is the same seam that runs through the district domain claim: the admin and legal scaffolding can be sound while day-to-day practice quietly leaks more than it should.

Nor does the designation resolve the question of enforcement. FERPA has famously weak teeth — its ultimate penalty is loss of federal funding, a sanction the Department of Education has never actually imposed. So the School Official status is best read as a set of binding commitments backed by contract and reputation more than by a regulator poised to punish. That's not nothing. But it's a floor, not a fortress, and the schools that treat it as a fortress are the ones most likely to get surprised.

Why this designation is the real story

Strip away the model names and the free-through-2027 headline, and the School Official clause is arguably the most important thing OpenAI committed to. Every district lawyer evaluating an AI tool asks the same first question: what's our lawful basis for the data this thing will see? "We designate ourselves a School Official with a legitimate educational interest, and the data stays under your control" is a real, recognizable answer in FERPA's own vocabulary — the answer that lets a general counsel sign off.

It also sets the bar for the competitive field. When a vendor names the statute and accepts the leash, it becomes harder for rivals to wave vaguely at "enterprise-grade security" without doing the same. That's the quiet function of using the exact legal phrase: it converts a trust question into a compliance question, which is the terrain schools actually know how to evaluate. The same logic underpins the union-brokered privacy work explored in why a union took AI vendor money — educators wanted the commitments written in language with legal weight, not slideware.

The clause won't make the hard calls for anyone. A teacher still has to decide what belongs in the chat box, and a district still has to write the policy that tells them. But by planting itself inside FERPA's School Official exception rather than beside it, OpenAI gave schools something concrete to hold it to. In student-data privacy, a designation you can point to in a statute beats a promise you can only read in a blog post.

Part 56 of 100 in the ChatGPT for Teachers series. Previously: The District Domain Claim, Explained. Next: Inside the $23M AFT-OpenAI-Microsoft AI Deal. Browse more builder insights or explore AI skills for education at aiskill.market.