SYNTHETIC IDENTITY DETECTION

Stop trusting the
green checkmark.

Your onboarding proves a face matches a document. It does not prove the face is a real human. Gradianza measures the authenticity gap your liveness stack was never built to see — interpretable, auditable, and independent of your verification vendor.

BOOK A FREE RETROSPECTIVE SCAN → ▸ Run the demo
NO INTEGRATION/VENDOR-INDEPENDENT/REGULATOR-DEFENSIBLE
ONBOARDING_SESSION // APR-4471
[ SUBJECT FRAME ]
KYC SCREEN✓ PASS
LIVENESS✓ PASS
FACE MATCH✓ 0.984
AUTHENTICITY ⚠ NOT MEASURED

Every gate above checks that the face matches. None checks that it is real. That blind spot is now your AML liability.

~A$1B
CBA-SELF-REPORTED LOANS TIED TO AI-GENERATED DOCS
2026
CRYPTO & VASP COMPLIANCE DEADLINES LANDING THIS YEAR
0
CONTROLS IN A FACE-MATCH STACK THAT TEST FOR REAL
01

Verification was never authenticity.

The dashboards say you're protected. Approvals look clean, liveness says "pass," the green checkmarks land. It feels like the controls are working.

Then the mule accounts and the suspicious-matter reports surface months later — and you realise the gate never checked the one thing that mattered.

This is not your fault. A synthetic identity sails through a face-match precisely because it was built to be internally consistent.

The whole industry sold "match" as if it meant "real." The generators caught up. Gradianza closes the distance between the two.

RETROSPECTIVE SCAN

Run it on a batch of "clean" approvals.

A sample queue your stack already approved. Every one passed KYC, liveness and face-match. Watch what Gradianza returns.

STATUS awaiting scan

Demonstration data. Your scan runs against your own recent approvals — no integration, nothing leaves an auditable trail you don't control.

02

A verdict you can actually audit.

[ 01 ]

Interpretable

Not a black-box score. Each verdict cites the synthetic signatures it found — frequency residue, texture entropy, generative artefacts — so a human reviewer and a regulator see the same evidence.

[ 02 ]

Regulator-defensible

Built for the question you'll be asked under AML/AUSTRAC scrutiny: how did you know? Every result is logged, reproducible, and defensible after the fact.

[ 03 ]

Vendor-independent

Gradianza sits outside your liveness stack. The verdict doesn't inherit your vendor's blind spots — it's a second, independent check on the one thing they never measured.

03

The market is already moving.

SOURCESIGNALSTATUS
CBA ~A$1B in suspected loans self-reported as tied to AI-generated documents. DISCLOSED
AUSTRAC Latest risk update names AI "identity fabrication" outright as an active typology. FLAGGED
VASP / CRYPTO Compliance deadlines for exchanges and virtual-asset providers land through 2026. 2026

The only question is whether you measure this gap before a regulator does.

FREE RETROSPECTIVE SCAN

Turn an invisible gap into a measured number.

If you run remote onboarding at a bank, lender, crypto exchange or VASP, we'll run a free retrospective scan of your recent approvals. No integration. You get a measured authenticity rate on identities you've already let through.

  • Runs against your own recent approvals — not a benchmark set.
  • Zero integration. No change to your onboarding flow.
  • A single defensible number you can take to your board.
REQUEST_SCAN.sh

Demo form — submits locally, nothing is transmitted. Live deployment routes to a secure intake.