You already have the governing authority. It just isn't enforceable yet.
Universities are not short of policy. The honour code, the IRB manual, the faculty handbook, FERPA — all of it exists and none of it reaches the moment a system acts. This scenario shows where it could.
What you are already obliged to produce
Institutional policy here has legal force. The gap is not authorship — it is enforcement at the point of action.
| Authority | What it obliges you to produce |
|---|---|
| FERPA §99.30 | Prior written consent before disclosing personally identifiable information from education records, with narrow enumerated exceptions. |
| FERPA §99.32 | A record of each disclosure, kept with the education record and available to the student for inspection. |
| IRB protocol | Human-subjects research confined to the approved protocol; secondary use of data is a protocol question, not a technical one. |
| Title IX | Procedural obligations in handling reports and records, including who may see what and when. |
Why the usual answer does not close
Policy written for people does not bind a system that never read it.
The honour code and IRB manual govern conduct, but nothing translates them into a constraint a research tool must satisfy before it acts.
Disclosure records under §99.32 assume a human made the disclosure. When a pipeline does, the record is often reconstructed afterwards from logs, if at all.
Departments adopt tools independently, so the enforcement surface is fragmented and the registrar has no single point at which to hold the line.
One action at the boundary
The compiler turns the policy you already approved into a deterministic constraint. It does not ask the department to write a new framework.
Candidate action
student.records.disclose · 340 transcripts · third-party analytics vendor
No consent basis and no enumerated exception was bound to the request. The disclosure is refused before it happens, and the refusal itself is written to the ledger — which is the §99.32 record the institution would otherwise be reconstructing later from application logs.
The same measurements, whatever the sector
These are figures from the internal technical evidence report, not projections modelled for this scenario. They describe one pipeline, so they do not change when the mandate does.
0 / 18
Attack episodes passed
Stage 0 SHADOW evaluation of action traces authored by real generative planners. None of nine held-out benign episodes was blocked.
0 / 10
Prohibited cases passed, dual reference
Controlled ablation on the same held-out set. A single-reference gate passed four of ten under identical calibration.
0.47 ms
Mean decision latency
Across 5,000 measured decisions; 1.15 ms at p99. The gate itself completes in 258 nanoseconds.
1,000 / 1,000
Identical digests on replay
Same input, pinned environment. Ten thousand ledger records re-verified in 0.195 seconds.
What this scenario is not
Read this before you quote it
- An illustrative scenario, not a customer engagement. No client is named because none is being described.
- No deployment in this sector is claimed, and no regulator has reviewed or endorsed this material.
- The figures are prototype measurements on one commodity workstation, not production or distributed results.
- Nothing here has been independently reproduced by a third party.
- The authorities cited describe the obligation you carry — not a determination that we satisfy them.
- Operational accuracy on a validated sector corpus remains outstanding; a pilot requires your own reference data.
Check the numbers before you trust the scenario
Every figure above is drawn from the technical evidence page, where the same measurements appear with their tail distribution and their stated limits.