Today, the system that produces a VIPS determination and the tool that verifies it both run inside infrastructure Vaxelrod controls. That is a fair question for any inspector general to ask, and the answer we are building is architectural rather than rhetorical.
The evidence vault, the signing authority, and the determination gate stay on one provider, under write-once retention, with a single point of issuance. Multi-cloud does not mean scattering the authoritative record across vendors. Splitting the record would weaken it, so we did not split it.
A verification job runs in a second cloud, in a separate trust domain, holding no credentials into the account that produced the determination and no privileged network path back to it. It receives only the exported package, handed over as a file, and re-runs a pinned verifier and a pinned proof toolchain against it.
If verification succeeds on infrastructure that cannot reach ours, the determination rests on the evidence and the mathematics rather than on trust in the vendor. The same job can be run by an assessor, an inspector general, or a prime contractor on their own infrastructure. That is the intended end state.
We did not put the sealed evidence behind another provider's analytics engine, and we did not let a probabilistic filter stand in for a deterministic gate. Services that would have required granting outside compute read access to sealed evidence were declined. The verifier's value comes precisely from what it is not permitted to reach.
Sealed intake, signed determinations, and independent verification of sealed records are running in production today. Verification in a second cloud trust domain is live under the v8.0 program.
Pilot engagements are open now: sixty days, fixed fee, one workflow. All initial conversations are under NDA.