Ungoverned AI | Regulation | Higher Education
The Accountability Gap: What Frameworks Built for Explainable Decisions Cannot See
By Dr. Freddie Seba · Wednesday, September 30, 2026
© 2026 Freddie Seba. All rights reserved.
Our miniseries on AI governance across highly regulated industries continues. Last week we examined healthcare; this week, higher education. I focus on healthcare, higher education, and financial services because I am a researcher-practitioner with hands-on experience in all three. The goal is not to prescribe. It is to surface lessons leaders can adapt to their own context and constraints.
The Opening Signal
In Issue 89, we saw governments move from voluntary AI commitments toward verifiable obligations. Higher education shows what that shift looks like on the ground.
In late April 2026, ShinyHunters breached Instructure’s Canvas learning management system. According to public reporting, the attack exposed:
- data from roughly 9,000 educational institutions
- names, email addresses, student ID numbers, and private messages
Instructure said it reached an agreement with the attackers and destroyed the stolen data. A class action and a congressional inquiry followed.
The breach did not involve AI. But it made one point unavoidable: campuses depend heavily on vendors they do not govern. The AI tools that now flag students’ work, rank applicants, and advise students run on the same dependency.
The board question is not only “Was our vendor secure?” It is “Who at our institution owned that risk?”
Question 1: What Were the Frameworks Designed to Govern?
Higher education’s core accountability frameworks were built for decisions people can make and explain.
- FERPA (the Family Educational Rights and Privacy Act) governs who may access and disclose education records. It gives students the right to inspect their records and to seek corrections.
- Civil rights law prohibits discrimination:
- Title VI covers race, color, and national origin.
- Title IX covers sex.
- Section 504 and the Americans with Disabilities Act (ADA) cover disability.
- Accreditation expects institutions to apply policies consistently and to show evidence of quality.
- Student conduct codes and institutional policy set out how discipline works. At public institutions, constitutional due process also applies.
Each framework assumes the same thing: a person made the decision, using a process that can be documented, explained, and reviewed.
Question 2: The AI Governance Gap — What Falls Outside the Frame?
AI tools strain that assumption in four distinct ways.
1. Validity. AI-writing detectors produce false positives.
- Weber-Wulff and colleagues tested more than a dozen detection tools. They found them neither accurate nor reliable (International Journal for Educational Integrity, 2023).
- Liang and colleagues found that GPT detectors misclassified more than half of essays by non-native English writers as AI-generated (Patterns, 2023).
2. Explainability. An AI detector gives a score, not a reason. Neither a faculty reviewer nor a student can challenge a number without an explanation.
3. Change over time. AI vendors update their models. A tool validated last year may behave differently today, often without the institution knowing.
4. Genuine variability. Chatbots and AI agents used for advising, tutoring, or student services can give different answers to the same question. That is how they are designed. But it means no single answer can be reproduced later.
The federal landscape is also shifting. On July 23, 2026, the U.S. Department of Education rescinded its regulations permitting Title VI disparate-impact claims. The Department now treats Title VI as prohibiting only intentional discrimination. Institutions should confirm the current status of earlier federal guidance on AI and discrimination, which relied on disparate-impact analysis. Disability law, state civil rights law, accreditation standards, institutional policy, and the courts remain.
When federal guidance recedes, accountability does not disappear. It moves to plaintiffs, courts, states, and boards.
Where It Breaks Down: Three Illustrative Scenarios
These scenarios illustrate how an audit or review could unfold. They are not findings from any actual regulator.
Scenario 1: The detection tool (Provenance)
Regulator: Show us the evidence that this detection tool is valid for your student population.
Institution: The vendor reports high accuracy.
Regulator: What is its false-positive rate for international/non-native English writers and for students with disabilities? Which model version was in use when this student was flagged? Did the student receive the report?
Institution: We would need to ask the vendor.
The board question: Can we document the basis for every AI-assisted determination that affects a student, including the tool version and its known limitations?
Scenario 2: The appeal (Processes)
Regulator: The student appealed. Show us the review.
Institution: A faculty member reviewed the flag and upheld it.
Regulator: On what basis? Did the reviewer know the tool’s false-positive rate? Did the student have a meaningful chance to respond before any sanction?
The board question: Is human review informed, or does it simply approve the tool’s output? A logged review that could not have reached a different outcome is not oversight.
Scenario 3: The advising agent (Product Ownership)
Regulator: Your AI advising assistant told two students different things about the same withdrawal deadline. One lost financial aid. Who is responsible?
Institution: The vendor provides the assistant. IT manages the integration. Student services uses it.
Regulator: Which one person is accountable for what it tells students?
The board question: Who, by name, owns each AI system from purchase to retirement, including what it does to students?
The AI Governance Documentation Gap
These frameworks assume an institution can explain why it made a decision affecting a student. AI detection tool scores and model outputs cannot supply that explanation on their own.
If no one can explain why a tool flagged a student, no one can review that flag independently. That is not a technology problem. It is an AI governance gap, and only AI governance as leadership can close it.
Question 3: What Does AI Governance as Leadership Look Like?
The question is not “Is our detection tool accurate?” It is “Have we built the minimum structures of accountability before AI decisions reach students?”
That is the heart of AI Minimum Viable Governance (AI-MVG), organized through the Seba 12 Ps of Responsible Oversight:
- Provenance: A documented basis for every AI-assisted determination, including the tool, the version, its known limitations, and the report the student can see.
- Processes: Informed human review, with escalation and reversal, before any consequence.
- Protections: Safe fallbacks when tools carry known false-positive risks for specific student groups, and disability accommodations built into academic integrity procedures.
- Product Ownership: One named person accountable for each AI system’s full life cycle. Not a committee, and not the vendor.
- Policy: Clear rules for when staff may use a tool, when they must not rely on it alone, and what triggers suspension.
- People: The students most exposed to false positives are part of the design, not an afterthought.
The Issue 89 AI governance readiness questions, applied to campus:
- Incident definition: Is a wrongful AI-based accusation treated as an incident, with a clear notification path?
- Stopping mechanism: Who can suspend a detection or advising tool today, without waiting for a committee?
- Independent verifier: Has anyone outside the purchasing office validated the tool on your students?
- Audit trail: Can you reconstruct what the tool reported, which version produced it, and who acted on it?
- Attesting executive: Would your provost sign a statement that your AI tools are used fairly and lawfully?
Next Steps for Boards and Presidents
- Inventory every AI tool that flags, ranks, advises, or disciplines students.
- Name a product owner for each tool: one person, accountable.
- Validate or suspend. Don’t use any detection tool that cannot be validated on your population alone to support a sanction.
- Align offices. Bring together academic integrity, disability services, IT, general counsel, and financial aid.
- Make AI governance a standing board agenda item. This is institutional risk and student rights, not IT infrastructure.
Download the AI Minimum Viable Governance Exercise: a two-page board tool to bring to your next meeting. [freddieseba.com/ai-minimum-viable-governance-exercise]
Friday: AI Enforcement in Higher Education, and what courts are deciding while federal guidance pulls back.
Let’s Continue the Conversation
Which AI tool on your campus has no named owner? Share it in the comments.
- For boards, presidents, and provosts: AI governance keynotes, board briefings, and embedded advisory built on AI-MVG and the Seba 12 Ps. Visit freddieseba.com or send me a direct message on LinkedIn.
- For researchers and institutions: I welcome collaboration on higher education AI governance research, convenings, and pilots.
- Subscribe: Ungoverned on LinkedIn and Substack.
References
- Instructure / Canvas breach: Inside Higher Ed, “Instructure Pays Ransom to Canvas Hackers” (May 11, 2026); Reed Smith, “Canvas/Instructure cyberattack: Key developments and action items for higher education institutions” (2026)
- U.S. Department of Education, Office for Civil Rights. Avoiding the Discriminatory Use of Artificial Intelligence (November 2024)
- Rescission of the Education Department’s Title VI disparate-impact regulations (July 23, 2026); Executive Order 14281 (April 2025)
- Weber-Wulff, D., et al. “Testing of detection tools for AI-generated text.” International Journal for Educational Integrity (2023)
- Liang, W. et al. “GPT detectors are biased against non-native English writers.” Patterns (2023)
- Ada Lovelace Institute and Alan Turing Institute. National survey of UK public attitudes to AI (2026)
- Seba, F. Ungoverned: A Practical Guide to AI Minimum Viable Governance (2026)
Disclosure and transparency. This analysis was created the way I argue AI should be used: critically, grounded in human judgment, for people, not in place of them. AI tools, including Anthropic’s Claude, OpenAI’s ChatGPT, and Google’s Gemini, support research synthesis, source verification, and drafting. Output may carry AI watermarks; a watermark confirms processing, not authorship. The frameworks, analysis, and editorial judgment are mine. Sources include government documents, academic research, policy analysis, and primary sources I read myself, analyzed with doctoral rigor and practical urgency.
Not legal advice. For information and education only. Consult your own legal counsel, compliance, and domain advisors.
#HigherEdAI #AIGovernance #FERPA #AcademicIntegrity #StudentRights #ADA #Section504 #AIMVG #Seba12Ps #BoardGovernance #HigherEducation
© 2026 Freddie Seba. All rights reserved.
