Penumbra Risk is the framework to quantify institutional exposure from partially-visible AI agents.
Between an agent's autonomous action and the institution lies a penumbra zone. Penumbra Risk turns that partial visibility into a clear metric.
The paper formalizes penumbra risk as a function of three variables — probability, consequence, and opacity — and proposes a verification protocol.
Publication awaiting peer reviewA governance checklist with PDF export, QR code, and Community Record. Currently available in English.
Open the toolIn collaboration — with invaluable technical insights from Ben Ogden
Ecuadorian attorney and legal-tech founder based in Washington, D.C. Daniela advises institutions on AI governance and institutional risk, and created the Penumbra Risk framework to quantify exposure from AI agents operating with partial visibility.
Ben provided the technical implementation of Penumbra Risk's tooling and governance checklist — and invaluable insights on the tool's implementation and design.