AICHE — AI Identity Council for Higher Education

AI agents are already on your campus.
Is your identity infrastructure ready?

This hub exists because higher education CIOs deserve a practitioner-built framework for managing AI agent identities — not a vendor brochure. Everything here was developed from primary research, real institutional experience, and the source material your identity team needs to get ahead of this.

Take the Survey Meet the Advisory Panel

"AI agents are writing emails, pulling student records, managing calendars, and making decisions without anyone watching. The question isn't whether to govern them — it's whether your institution gets ahead of it or gets caught by it."
— Raymond Todd Blackwood, President, QuickLaunch

What's Actually Happening on Campus

The problem is already inside the building.

Most institutions are managing AI agents the same way they managed applications in 2005 — shared credentials, static API keys, and zero separation between what a human does and what an autonomous agent does. That's not a gap on the horizon. It's a gap that's open right now.

Shadow AI Is Already There

Departments are spinning up AI agents without IT approval. If you don't have a registry, you don't have visibility. If you don't have visibility, you don't have security. The attack surface you can't see is the one that gets you.

Your Vendors Are Shipping Agentic Features Right Now

Check your release notes for three terms: "Model Context Protocol (MCP)," "Dynamic Client Registration," and "Rich Authorization Requests." If those phrases don't appear in your current vendor's roadmap by Q3 2026, that's a conversation you need to have today.

Static Secrets Are Your Biggest Unaddressed Debt

Long-lived API keys pasted into configuration files, Jira tickets, and GitHub repos. One leak means full agent compromise — and agents operate at machine speed, not human speed. A compromised agent can exfiltrate millions of records before anyone notices.

What Good Looks Like

A framework organized by institutional context.

There is no one-size-fits-all answer here. An R1 research university with a dedicated security team has different options than a community college with two people covering all of IT. Start where you are, not where the enterprise security vendors assume you are.

⚙ Design

Implement a Zero Trust architecture that treats every AI agent as a unique, non-human identity (NHI) with its own credentials. Access must be Just-In-Time (JIT) and ephemeral — privileges granted for a specific task, then immediately revoked. Agents operate within strict sandboxes and can only interact with external services listed in an approved tool registry.

⚖ Governance

Define acceptable use policies that cover AI agents explicitly — not just human users. Establish a formal agent registration process before any agent touches institutional data. Include AI agent scope in your existing data governance and FERPA compliance reviews.

🔎 Monitoring & Auditing

Deploy Identity Threat Detection and Response (ITDR) to catch real-time policy bypasses — skipped MFA, privilege escalation, anomalous call patterns. Maintain immutable, tamper-proof logs that record agent activity entirely separately from human user activity. These logs are your forensic foundation if something goes wrong.

⚙ Design

You don't need to rip out your existing infrastructure. Use an Identity Fabric approach — integrate your existing directories, user stores, and identity providers rather than replacing them. Stand up a centralized API gateway as your security choke point. All agent traffic routes through it. Security logic lives there, not scattered across individual application configurations.

⚖ Governance

Implement a lightweight agent registry — even a spreadsheet-tracked approval process is better than nothing. Require department heads to submit AI agent requests to IT before deployment. Document what each agent is allowed to access, at what scope, and for how long.

🔎 Monitoring & Auditing

Turn on separate audit logging for agent activity in whatever IAM platform you're running. If your current IDP can't separate agent logs from human logs, that belongs on your next vendor review agenda. Rotate static credentials on a minimum 30-day cycle as your interim posture while you work toward dynamic credentials.

⚙ Design

Start with discovery, not infrastructure. You cannot secure what you cannot see. Use Identity Security Posture Management (ISPM) — or even a basic network scan — to find what's running before you spend a dollar on new tooling. Your first 30 days should be observation only.

⚖ Governance

Adopt the 90-Day Inspect-Protect-Govern sprint below. No new budget required for Phase 1. Build your agent registry as a simple form-based intake process. The goal is awareness first, enforcement second.

🔎 Monitoring & Auditing

Minimum viable audit posture: know which agents are running, what credentials they're using, and who is accountable for each one. A named human owner for every deployed agent is a governance control that costs nothing and provides significant accountability.

Inspect. Protect. Govern.

A 90-day sprint for any institution.

Designed for institutions with no dedicated AI governance team and a constrained IT budget. This is a sprint, not a transformation program. Three phases. Ninety days. No rip-and-replace required.

Days 1–30

Inspect

Deploy ISPM tooling or run a manual discovery audit to identify all AI agents, shadow AI models, and AI-enabled features currently operating in your environment.
Inventory all static API keys — search GitHub, Jira, wikis, and configuration files for hardcoded credentials. Every one you find is a risk you can now address.
Map which existing vendor platforms (ERP, LMS, IAM) have shipped agentic capabilities in recent release notes, whether or not anyone in your institution turned them on.

Days 31–60

Protect

Stand up a centralized secrets vault. Pull hardcoded credentials out of applications and into managed, rotatable storage. This is the single highest-ROI security action you can take with minimal spend.
Implement an API gateway to handle token exchanges. This centralizes your security enforcement and instantly improves the posture of legacy applications.
Begin migrating agents from shared human credentials or static service accounts to dedicated agent identities with scoped, minimum-necessary permissions.

Days 61–90

Govern

Publish an AI agent acceptable use policy. Define what agents are permitted to access, what data they may process, and what approval process is required before deployment.
Establish a vetted tool registry — a list of approved APIs, databases, and external services that agents are authorized to call. Anything not on the list requires a review.
Enable independent agent audit logging. Ensure agent actions are recorded separately from human user activity. Run your first audit report before Day 90.

Three Questions. Three Disqualifiers.

What to ask your identity provider.

Before your next vendor renewal or IDP evaluation, ask these questions. The answers tell you more than any feature sheet.

Question 1

Do you treat AI agents as distinct identities — separate from human users and traditional application accounts?

✓

What you want to hear

Yes. Agents have their own registration process, credentials, permissions, and audit trail — completely separate from the human identity directory.

✗

Red flag

"Agents use service accounts" or "we recommend a dedicated user account for each agent." That's not agent identity. That's a workaround with a compliance gap built in.

Question 2

How do you handle Model Context Protocol (MCP) authorization?

✓

What you want to hear

The Authorization Server is fully separated from the MCP Resource Server. The IDP acts as a standalone auth layer — it does not also serve as the data resource endpoint.

✗

Red flag

Any answer that describes the authorization and resource functions in the same component. Collapsing these creates architectural vulnerabilities that are difficult to audit and nearly impossible to cleanly remediate at scale.

Question 3

Do you support OAuth 2.0 Token Exchange for multi-agent flows?

✓

What you want to hear

Yes. When an agent calls a downstream agent or system, it must exchange its current token for a new, narrowly scoped token at that node. Identity is re-proven at every hop.

✗

Red flag

A single token passed through the entire multi-agent flow. If an agent's token is compromised at any point in the chain, an attacker has access to every downstream system that token touches.

Source Material

Reading Room

Everything referenced in the AICHE framework is linked below. Primary sources, not summaries. Read the originals.

PDF

The AICHE Advisory Panel

Practitioner voices. Not vendor voices.

The AI Identity Council for Higher Education brings together technology leaders from institutions across the country to define practitioner standards for AI identity governance in higher education. Panel members contribute directly to the 2026 Field Guide and participate in working sessions that shape the research findings.

This is not a vendor advisory board. It is a practitioner body. QuickLaunch sponsors and publishes the research. The panel owns its voice.

Panel Forming Now — Announcement April 2026