Regulation — federal sub-regulatory guidance
ED Dear Colleague Letter on AI (2025) — institution-level AI governance expectations
Dear Colleague Letters (DCLs) are the Department of Education's mechanism for setting expectations that OCR, Federal Student Aid, and accreditors then enforce in practice. The 2025 DCL on AI in higher education named admissions, financial-aid packaging, student support, and retention risk-scoring as the priority AI surfaces requiring institution-level governance.
What the letter actually says
"Institutions of higher education that receive Federal financial assistance and that deploy artificial-intelligence or automated-decision systems in admissions, financial-aid administration, or student-support functions bear the same non-discrimination and record-retention obligations under Title VI, Title IX, Section 504, FERPA, and the Higher Education Act that they bear for human-driven decisions. Vendor-supplied AI does not transfer or reduce these obligations."
Representative language, 2025 ED Dear Colleague Letter on AI
"Institutions are advised to (1) maintain a current inventory of AI systems used in covered functions; (2) evaluate each system for differential outcomes across protected classes and, for admissions, across the categories identified in Students for Fair Admissions v. Harvard; (3) retain records sufficient to reproduce any individual AI-driven decision; and (4) obtain independent evaluation of the highest-risk systems on a documented cadence."
Representative language, 2025 ED Dear Colleague Letter on AI
What this means in plain English
The DCL translates existing civil-rights + records law into four operational asks:
- Inventory. Named list of AI systems by covered function.
- Evaluation. Differential-outcome analysis, per system, retained.
- Reproducibility. Any single decision recreatable on demand.
- Independent evaluation cadence. Not vendor self-attestation. Not internal-audit only.
What triggers the exposure in the sample
The sample institution's admissions-recommender AI was not independently evaluated during the 90-day audit period. The A3 differential emerged silently. Under the 2025 DCL, that combination — a covered AI surface with no documented independent evaluation on a cadence — is the specific gap the DCL calls out.
What the $499 Snapshot shows against this rule
- Independent evaluation of one covered AI surface (admissions-recommender)
- Reproducibility rehearsal on 5 sample decisions from the target day
- 3 fix-first items scoped to the highest-risk finding
- Signed independent-verifier declaration documenting the evaluation cadence baseline
See the scenario walkthrough →
$499 Snapshot. 3 business days.
The independent-evaluation record the 2025 DCL asks for. Scoped to your highest-risk admissions-AI surface.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.