The AI profile on each system says whether it is AI-assisted or agentic. This inventory goes one level down: each model, feature and agent, its risk tier, who reviews it, the data it can touch and what it can change.
Flagged automatically from the inventory: high-risk uses without a system card or without a person approving each output, and agents that can change sensitive or restricted data.
{{ f.reason }}
Showing {{ shown }} of {{ count }}. Select a row for its oversight and blast radius.
Shapes a decision about a person's admission, grades, standing, employment, health or access, or can act on restricted data.
Talks to people or produces content they see, or takes bounded actions a person can reverse.
Internal operations with no decision about an individual and no restricted data.
The tiers follow the shape of the EU AI Act and NIST AI RMF without claiming a legal classification. Data reach uses the same four classification levels as the catalog.
An entry here is the index. The system card in the registry is the record.
Edit ai-inventory.json as uses are proposed, piloted and retired. Each entry should point to a system card in the Campus AI Registry.