Regulation — federal privacy + access
FERPA (20 USC 1232g) — AI-decision record retention + access rights
The Family Educational Rights and Privacy Act governs student education records, including admissions files. When AI drives an admissions decision, the model version, input snapshot, and decision-hash all become part of the education record. FERPA sets both the retention floor and the access-request duty.
What the regulation actually says
"No funds shall be made available under any applicable program to any educational agency or institution which has a policy of denying, or which effectively prevents, the parents of students … the right to inspect and review the education records of their children."
20 U.S.C. §1232g(a)(1)(A) — the access-rights hook
"The term 'education records' means, except as may be provided otherwise in subparagraph (B), those records, files, documents, and other materials which — (i) contain information directly related to a student; and (ii) are maintained by an educational agency or institution or by a person acting for such agency or institution."
20 U.S.C. §1232g(a)(4)(A) — the definition that captures AI-decision records
What this means in plain English
Three things for AI-driven admissions:
- The AI decision-record is an education record. Score, lane routed, model version, input snapshot — all fall within FERPA's definition once the applicant becomes a student.
- The parent or eligible-student can request it. Denied applicants who later matriculate elsewhere may still surface the records via litigation discovery or state-level analogs.
- Retention is not optional. Institutions must maintain records so that access requests and OCR investigations can be answered on the schedule the request demands.
What triggers the exposure in the sample
Approximately 558 A3 applicants over the 45-day drift window were potentially misrouted. Each one produces a FERPA-accessible AI-decision record. If any single one requests inspection under 20 USC 1232g, the institution must produce — the model version, the input snapshot, and the score at decision-time. Without a retention discipline in place, that production is expensive at best and adverse-inference-generating at worst.
What the $499 Snapshot shows against this rule
- 5-decision counterparty-question rehearsal that proves reproducibility on demand
- Model version pinned + decision hash bound + input snapshot retained per decision
- Retention pipeline independent of the production admissions model
- Signed independent-verifier declaration attesting the retention discipline was operational as of the Snapshot date
See the 5-decision reproducibility rehearsal →
$499 Snapshot. 3 business days.
FERPA-grade retention rehearsal on your institution's actual admissions-AI decision stream.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.