Evidence with citations
Every screening finding is grounded in a source document and list entry. Reviewers and auditors can trace any claim back to where it came from - not to a model's recollection.
A governed AML/KYC onboarding run, replayed end to end: source-grounded evidence, deterministic screening checks, then a human approval gate where a compliance officer rejects a confirmed sanctions hit and writes a tamper-evident proof packet.
Read-only replay of an illustrative run - no inputs, no production credentials.
A real governed run, replayed in full below - ending in a human rejection and an auditable proof packet.
Eight events, exactly as the platform recorded them - from intake to a human approval gate that stops onboarding. Nothing is simulated past what you see here.
A new-customer onboarding application was submitted for governed screening. The run is registered against flow aml-kyc/v4 and begins under the workspace's controls - every step from here is recorded.
Three documents were parsed and source-grounded evidence was extracted - each finding carries a citation back to the document and screening source it came from.
Rule-based checks ran against the screening rule set. The identity checks passed - but a confirmed sanctions match fails the gate later depends on.
Workspace policies were evaluated against the extracted evidence. One policy trigger fired on jurisdiction exposure.
The findings were combined into a composite onboarding risk score. The score is high, so the run is routed to a compliance officer rather than auto-cleared.
Above the 70 review threshold -> routed to a human approval gate. No account is opened automatically.
A compliance officer examined the run and rejected it at the approval gate. The customer cannot be onboarded, and the run stops here - no account is opened and no downstream write is performed.
The run ended exactly where governance required: a person decided, the decision is logged, and there is a proof packet anyone can audit later. No account was opened on a confirmed sanctions hit.
The replay is not a chatbot demo. It is the control plane doing its job: grounding, gating, and proof - the parts a regulator asks about.
Every screening finding is grounded in a source document and list entry. Reviewers and auditors can trace any claim back to where it came from - not to a model's recollection.
A confirmed sanctions hit routes to a compliance officer, not an auto-approve. Here the officer rejected - and rejection is final at the gate. The run stops before any account is opened.
The decision, the officer, the reasons, the SAR basis, and the cited evidence are written once to an append-only audit trail. Months later, you can prove exactly why this applicant was rejected.
PrivateFlow is not certified under any compliance framework. Controls are designed to support compliance preparation. This run is an illustrative, synthetic example.
We'll stand up a governed pilot on your own data - sandbox first, no production credentials required. You decide what the gates do.