Financial Services
Check claim photographs and recorded claimant statements for signs of editing or AI generation, and see the evidence behind every finding
Overview
Check claim photographs and recorded claimant statements for signs of editing or AI generation, and see the evidence behind every finding.
An insurer sees the damage through a photograph somebody else took, and hears the account of what happened through a recording. Neither can be examined in person, and both can now be produced by software convincingly enough that looking harder does not settle it.
The pressure runs in both directions. A claim refused on a hunch costs a genuine customer and invites a complaint. A claim paid on a fabricated photograph is rarely recognised as a loss at all, because nothing on the file looks wrong.
The checks below examine what was submitted and return a finding with the reasoning attached, so an assessor can act on it while the claim is still open and explain that decision later to the customer, an auditor or an ombudsman.
What you deploy
These are the FaceOff products most often deployed in this sector, in the order they usually land.
Key capabilities
What this sector asks for most often. Each one runs at the moment a decision is being made, and each returns a result a reviewer can explain to somebody else.
Images attached to a claim are checked for signs of editing or AI generation, and the finding points to the areas of the picture that look altered. An assessor sees what was noticed rather than being handed a verdict to take on trust.
Recorded video from a claimant is checked for face swaps and generated footage, and the finding points to the moments worth watching. A long file becomes a short list of places to look rather than an afternoon of playback.
Some policies and the claims made against them are built from genuine details belonging to different people, which is why every detail checks out on its own. Records are read as a connected group instead, so a set of policies that no unrelated customers would share becomes visible.
No finding arrives on its own. Each carries what was noticed and where, so an assessor can act on it now and the decision can still be explained months later to the customer, an auditor or an ombudsman who has asked why the claim went the way it did.
Some claims are not one doctored document but an impersonation carried through the whole submission. The attempt itself is assessed and returned as a graded result, so a borderline claim can be routed for closer handling instead of being waved through or refused outright.
Most claims are exactly what they appear to be, and the checks are meant to pass over them. A claim that raises nothing continues through your normal process, so attention lands on the small number that earned it rather than on every customer who has had a bad week.
Where it is used
Every entry names a product on this site and describes it the way its own page does.
DeepImageGuard checks a submitted image for signs of editing or AI generation, and shows which areas look manipulated.
DeepVideoGuard checks recorded video for face swaps and generated footage, and points to the moments worth reviewing.
DeepFraudGuard assesses the risk that a claim is a deepfake impersonation or an AI assisted attempt, and returns a graded result.
SyntheticFraudGuard AI reads identity records as connections, surfacing groups built from stolen and invented details.
Common questions
Yes. DeepImageGuard checks a submitted image for signs of editing or AI generation and highlights the areas that look manipulated, so an assessor can see what was found rather than rely only on a score.
Yes. DeepVideoGuard checks recorded video for face swaps and generated footage, and points to the specific moments that need review rather than asking someone to watch the whole file.
No. It returns a graded finding with the evidence behind it. The decision on the claim stays with your assessors, following your own process.
On the material as it is submitted, which means a finding is available while the claim is still open and before it is settled. That is the only point at which it can change the outcome; a check that arrives after payment has been made is a record of a loss rather than a way of preventing one.
These checks answer whether an image was generated or altered, not whether it has been submitted before, so an unedited photograph reused from an earlier claim is not what they are looking for. Duplicate submissions are a question for your claims system. Where the same people and details keep recurring across claims that ought to be unrelated, SyntheticFraudGuard AI reads those records as connections and surfaces the group.
Compressing, resizing and forwarding a file are all normal in claims handling and are not in themselves signs of tampering. What comes back is what was found in the file that was given, together with the areas or moments behind it, so an assessor can weigh a finding on a heavily forwarded image differently from the same finding on an original.
No. It sits alongside them and answers a question they were not built for, which is whether the photograph, the video and the person in front of you are genuine rather than whether a claim looks unusual against your historical patterns. Findings come back in a form your existing assessment, case and audit processes can carry without a new workflow.
Analysis can run inside your own environment where claim material is sensitive or must stay in the country, so checking a claim does not require handing the photographs and recordings of your customers to somebody else in order to protect them.
Any method of detection can be mistaken in either direction, human judgement included, so a finding is an evidence based assessment rather than a guarantee. That is exactly why the evidence comes with it. It is there to focus an assessor on what deserves scrutiny and to be examined and challenged, not accepted without question.
Bring your own footage. We will run it through the Adaptive Cognito Engine and walk you through the findings.
