A motor claim arrives with photos of a crushed bumper and a repair estimate on a real garage’s letterhead, and the payout goes to an account that has received payouts from other claims. DeepID reads the camera the video came through, the device behind the claim, the documents for edits and the payout account’s links, so clean claims are paid without an extra step and the fabricated ones reach your investigators with the evidence attached.
Hold
HOLD · device shared · document tampered · payout account sharedA claimant opens the insurer’s app and files a motor claim. The photos show a crushed bumper and a cracked headlamp, the repair estimate carries the letterhead of a real garage, and every field matches the policy. An assessor reviews the file on a screen, sees nothing wrong and approves it. The payout goes to an account that has already received payouts from other claims.
An illustrative scenario, not a real customer’s case.
A ring generates damage photos, or edits a genuine hospital bill, certificate or repair estimate. The result carries real letterheads and details written to match the policy.
Your checks saw
Nothing yet. No claim filed.
DeepID reads
Nothing yet. Edits to a document leave traces that are read once it is uploaded.
The claimant uploads the media or completes a video self-inspection. A ring may feed prepared video into the camera from an emulator and file several claims from one device.
Your checks saw
An active policy. A complete claim form.
DeepID reads
A virtual camera on an emulator, and a device that has filed other claims.
A handler checks the documents by eye and against policy data. With no view of the device or the linked claims, the file looks clean.
Your checks saw
Documents consistent. Within policy cover.
DeepID reads
Supporting documents checked for edits, and the claim linked to repairers and payout accounts on other claims.
The same account, or accounts tied to the same people, collects payouts from claims filed under other names.
Your checks saw
Payee details supplied by the claimant.
DeepID reads
One hold before payout, with the device, document and payout-account reasons.
Each of these trusts the picture it is given. Read together across the journey, they stop agreeing, and that is what DeepID reads.
Generated photos and edited bills now look genuine to a trained eye on a screen.
The ring writes names, dates and amounts on the fabricated documents to match the policy record.
A prepared video fed into the camera feed looks like a live capture.
Each claim is filed under a different identity, so the history for any one claimant looks clean.
Liveness on the self-inspection video and the claimant’s face check catches injected camera streams and replayed video, with a signed record of the capture. It tests the person and the capture, not the damage.
Bills, estimates and certificates read and checked for edits as supporting documents. A letterhead can be real and the figures changed.
Emulators and virtual cameras read at the point of capture, and a device ID that survives reinstalls. A device that has filed claims under several names is something a single claim file never shows.
Claimants, repairers and payout accounts linked across claims. A payout account or a repairer that recurs across unrelated claimants turns a set of plausible files into a visible ring.
The payout account checked against the name on it before the money goes. It passes for a mule with a real name, which is why it is read with the graph.
Allow clean claims through with nothing added, step up when the capture cannot be trusted, and hold the payout when the device, the documents or the payout account link to other claims. Your handlers and investigators decide the claim; DeepID attaches the evidence.
A real camera, a clean device, documents without edits and a payout account with no links. The claim goes through with no extra step from DeepID.
The camera feed cannot be trusted, or the media quality is poor. The claimant films the damage again in a guided in-app capture with liveness.
A document shows edits, or the payout account is linked to other claimants. Payment waits while a handler checks the linked claims, with the flagged pages shown.
A virtual camera, a device shared across claims and a shared payout account appear together. The payout does not go, and the claim reaches your investigation team with the evidence and the linked claims attached.
“Your claim is with a claims handler, who will contact you before anything is paid.”
The same decision reaches your FRM, case manager or SIEM as an event, with one code per reason. The codes here are illustrative; real codes and their names are agreed in the sandbox.
Agreed in writing before the pilot starts, and counted on your own traffic against what you see today. Devices, documents and payout accounts can be measured on your own closed claims first, in a free retrospective. The capture path is proved in a shadow proof on live claims. We publish no figures for these.
It does not claim to. Detectors for generated images change as the tools change, so DeepID reads what is harder to fake: the camera the video came through, the device behind the claim, edits to the documents and the links between payout accounts. A retrospective on your own closed claims shows what those find.
No. DeepID holds the payout and explains why; your handlers and investigators decide the claim. Every decision is logged with its reasons, so the file shows what was found and when.
Device and camera signals need the SDK in the capture journey, so claims that arrive by email or through an agent carry less evidence. Document checks and the payout-account links still apply.
No market on the rule map has a rule written for claims evidence, and the insurance regulator’s fraud framework in India is not mapped yet. The controls below are the risk measures each market page dates and sources.
Device, behaviour, scam and payee signals read at login, new payee and payment, returned as one decision with its reasons.
DeepID supports compliance with these rules. It never claims to guarantee it.
Sign-up or login end to end, with the decision back in the app and as an event, so your engineers see the shape of it before anyone signs anything.