Risk Mandate maps every autonomous system to its blast radius, assigns a business owner, and drives a time-bound decision — accept, fund, or fix.
↑ Interactive — click a risk record, then act on it
Credentials, data, infrastructure, vendors — each has an owner, a register, and a decision process. Autonomous systems act with real authority and answer to none of them. Risk Mandate is the business risk layer for autonomous systems: map, accept, fund, and reduce.
Severity isn't in any one finding — it's in the combination. When a single agent holds all three, untrusted input can drive a privileged action straight to an external destination.
Anyone can list your agents. The work that matters is what comes after the list — and that work is a risk decision, not a longer report.
Discover every agent and trace what it can actually reach — systems, data, secrets, external destinations.
A named owner accepts the residual risk for a defined window. No deny button — accept, and revisit on expiry.
Acceptance manufactures budget and ownership: to keep running, someone funds the reduction.
Every fix has its own blast radius. The goal is to leave the system measurably safer — sometimes a control, not a code change.
Most agent "controls" are intentions the stack can't enforce. Risk Mandate separates what's actually enforceable from what's merely hoped — then puts a clock, an owner, and a budget on the gap.
How long, who owns it, what's funded.
Standing power nobody remembers approving is how agents become dangerous. A capability certificate grants the minimum required access, names the owner and justification, and auto-revokes on expiry. Authorisation becomes a fact, not a hope.
Response within one business day · No commitment.
Pre-launch. The thesis and model for a service being built — not a shipped product yet.