Risk is not beside the system — it governs it. Every consequential action, from any division, passes Risk's checks before it is authorized to move. When a check fails, the answer is a veto, in the open.
Exposure, limits, and the authority to hold anything that fails them — with the failing check quoted, every time.
A silent override. A veto is visible, recorded to memory, and reversible only by a human with the authority to sign for it.
Every division runs the same 7-Stage Governed Decision System, under the same Safety Guardrail Engine, writing to the same Compounding ROI Memory. Here is Risk's slice of it — stated plainly, so you can hold us to it.
Not a mockup — the live desk runs the production runtime, seeded and replayable, and this is one of its decisions.
Mid-run, an action models outside its declared limit. The desk stops it cold: the veto holds, the failing check is read out verbatim, and the hold itself is recorded to Compounding ROI Memory — the desk remembers why it said no.
Illustrative scenario data · deterministic and seeded — replay it yourself on the live desk.
The desk narrates its own run: the checks, the gates, and the moment it commits or holds. Same runtime that ships.