The Business Verification Playbook: KYB for the Prosecution Era

The threat model: four fictions, one economy

Corporate-vehicle fraud runs on fictions layered until scrutiny gives up, and four cover the field. The fictitious company: entities that exist only as filings, typically dissolving inside 200 days after moving sums their trading history cannot explain. The nominee human: a real, verifiable person whose name is on the filings and whose control is a private agreement the registry never sees. The obscured owner: control routed through layering, cross-holdings, trusts, and disclosure-light jurisdictions until no natural person appears above the threshold. The stale file: a KYB file true at onboarding and refreshed annually, defending a snapshot against a motion picture.

The economy behind all four is rational: a shell costs less than a synthetic identity, a nominee costs a fee, and the expected penalty was, until this month, an administrative letter. The convictions repriced the defense's side of the ledger. This playbook reprices the attack's.

Reference architecture: person-anchored KYB

The architecture that survives the threat model treats entity data as the outer shell and verified humans as the core, with evidence assembled by the system rather than the analyst. Two properties matter most. Verticality: every check feeds the evidence vault directly, so the case file assembles itself as the case happens. And symmetry between onboarding and monitoring: the same resolution and proofing machinery runs at account opening and on every triggering event afterward, so the back book ages at the same standard as the front door.

The program expresses this as four gates a business customer passes and keeps passing. The entity gate: the company is real, registered, and coherent with its declared business. The person gate: every director and UBO is a liveness-verified human bound to their claimed role. The control gate: the resolved structure matches observed control on the live account. The continuity gate: events re-open the earlier gates automatically, so passage is a state, not a ceremony.

Phase 0: book census and exposure audit (weeks 1-2)

Programs inherit books, and the book is where enforcement will look first. Phase 0 answers four questions per corporate customer: when was the entity last verified against its registry; which directors and UBOs are identified; which of those persons were ever identity-proofed rather than database-confirmed; and what evidence exists per answer. The output is an exposure register ranked by the product of gap and stakes, and it becomes the first exhibit of program good faith.

Phase 1: entity truth and resolution (weeks 3-5)

Phase 1 automates the outer shell. Registry verification runs against primary sources across every jurisdiction the book touches, with document forensics on the instruments the registries do not hold. The resolution engine assembles ownership across layers and borders until natural persons emerge or the dead end itself becomes a finding: an unresolvable structure is a risk verdict routed to enhanced due diligence, not a data-quality nuisance. Mismatch between registry, declaration, and observation is signal, logged and adjudicated, never silently reconciled to whichever source unblocks the funnel.

Phase 2: the person gate (weeks 6-9)

Phase 2 is the program's center of gravity: every resolved director, PSC, and UBO completes identity proofing as a person, not a record, through government document authentication, biometric matching, and structural passive liveness with injection defense, run from the person's own phone in minutes. The corporate twist is role binding: the verified identity is bound to the claimed position and cross-checked against the resolved structure. Two attack classes make the biometric layer non-negotiable: the rendered owner (a deepfake face presenting a genuine document on a remote call) and the industrial nominee (the same face or document recurring behind formally unrelated entities, caught by cross-entity reuse analytics).

Phase 3: the control and continuity gates (weeks 10-12)

Phase 3 makes passage a state. Event feeds wire into the decision plane: registry filings and director changes, ownership transfers, sanctions and PEP updates, adverse media, and the customer's own transaction posture where it contradicts the declared business. Each event class maps to a proportionate re-check within hours, not at renewal. The control gate adds the comparison legacy programs never run: resolved structure versus observed behavior, which is what catches the nominee arrangement the registry cannot see, because private agreements leave public footprints in behavior.

Phase 4: evidence and audit operations (weeks 13-16)

Phase 4 assumes the examination and builds backward. Per-customer case files export on demand: the entity's verification history, the resolved structure with its sources and dead ends, every person's proofing record with method and date, and every event trigger and its disposition, assembled in minutes. Two service levels are worth committing to in writing: any regulator request for a customer's complete verification evidence is met within 24 hours, and any confirmed sanctions touch on a resolved person reaches decision-makers the same day.

Phase 5: adversarial assurance (ongoing)

A KYB program that has never been attacked by its own side is a hypothesis. From week 16, a standing red-team calendar runs synthetic shell applications, nominee personas recycling faces across applications, rendered-face proofing attempts, quiet control transfers mid-quarter, and the stale-file drill. Each drill scores catch rate, time to detection, and evidence quality, and each miss converts to a dated fix.

Pull quote

“Data-layer vendors pass the paperwork; person-layer stacks catch the human, and the difference is the entire post-conviction era in one demo.”

— The buyer's evaluation test

The regulatory and liability map

Read as one map, the direction is uniform: registers centralize, thresholds harmonize around 25 percent with control-by-other-means, and enforcement moves from entity disclosure to person verification, with the UK simply arriving first at the courtroom.

Measurement and the program checklist

Ten numbers run the program, but two moving in opposite directions tell its health: proofed-person coverage rising, oldest-evidence age falling. When both stall, the funnel veto is back, and it is a leadership conversation, not a tooling one.