FOR EDUCATION
A STUDENT RECORD THAT SHOULD NOT MOVE DOES NOT MOVE.
Enrollment, records, scheduling, parent emails. AI is picking up work your staff never had time for, and it touches student data to do it. Mountain Theory checks every action your district's AI tries to take, against rules you write in plain English, before it happens. Nothing has to be caught afterward, because it never ran.
The question your board is about to ask
Districts are putting AI into classrooms, operations and staff workflows before governance standards exist. A grade changed by an assistant that misread a request. A class list pasted into a public model to draft a letter. A student record sent outside the district because a parent's address matched.
None of those is a breach. FERPA reads the same way whether a person or an AI moved the record, and the board question is not "do we have an AI policy." It is "what stops the action when the policy is wrong or the model changes."
How personal data is stopped at the action
What we do
We control what autonomous AI does. The model still decides. The action does not run until it has been checked.
- Stops tool chaining: blocked once, blocked again on the workaround
- Catches agent drift: the goal it was given, not the one it wandered to
- Zero-day agent behavior stopped the day it appears
- No new rule, no signature, no patch to do it
Personal data is checked in and out: student records, parent contact details, health and IEP data, payment data. Stopped before it leaves your environment, whether it was typed, pasted, or pulled from a system the assistant was allowed to read. Every action gets one of three outcomes, ALLOW, HOLD, or BLOCK, and every decision is logged with the rule that matched it. It works with whatever AI you already use: Gemini, Copilot, ChatGPT, Claude, or a custom build.
The scenario
A vendor updated the hosted model behind one of our own agents overnight, and the agent began reaching for the network to answer questions it used to answer from local records, improvising commands instead of the sanctioned scripts. Every attempt was stopped the day it appeared, with no new rule on our side. In the ungoverned lane of the same run, every one of them completed.
Your district's model changes on the vendor's schedule, not yours. The rules have to hold on the day it does.
The July run, recorded, with the two actions we had no policy for
One rule, as your records officer would write it
Policy
Sending a student record outside the district or deleting a record is prohibited for every agent. Changing a grade or exporting a class list goes to HOLD for a named person. Log every attempt.
Rules sit in one place and apply everywhere the AI runs, whichever assistant or model is behind it. Change a rule once and the change is live. Nobody waits on a vendor to do it.
What you hand the auditor
An append-only record of every decision: the action proposed, the rule that matched, the outcome, the time, and who wrote the rule. It exists because enforcement produced it, not because someone reconstructed it after a parent complaint. That is what a FERPA review asks for when an AI system touches student records, and it lets a district answer for its own AI use without a central team rebuilding the story later.
What the record has to contain, and why the usual logs do not count
Proof in this lane
A Tier 1 public research university in Texas has agreed a paid POC and pilot with Mountain Theory. The Recorded Run Protocol lets any institution check our claims for itself, misses included, before signing anything.
The Recorded Run Protocol, free to use with or without us
The ask
A 30-day paid proof of concept on your workflows and your action set, in your environment, not ours. Then one flat annual price for the district or institution, with no per-agent count. Tests published, misses included, before anyone signs.
Proof