Insurance
First notice of loss is a photograph. Proof of loss is a document. Both are now trivially generative, and both arrive in your claims system as attachments no adjuster can forensically evaluate at line speed. TruthScan scores them before reserve is set.
THE THREAT
The staged-accident economy was always limited by physics — someone had to actually dent the car, actually flood the basement, actually lose the laptop. Generative editing removes the physical act entirely: the claimant photographs an undamaged vehicle and asks a model to add the damage. The result is the real car, the real plate, the real driveway, with a dent that never existed. It is the hardest class of image fraud to catch by eye, and it is the most common one we see.
In seconds
End-to-end verdict, fast enough to sit inside straight-through processing rather than beside it.
In-paint
Localized manipulation detection — the specific failure mode of a real photo with fabricated damage added.
Heatmaps
Region-level evidence an adjuster can attach to a denial and an SIU referral can build on.
ATTACK PATTERNS
Written the way an SIU investigator would describe them, because that is who will read this page.
The claimant takes a photo of their undamaged bumper and edits it to show collision damage. The rest of the image is authentic, including the plate and timestamp. Metadata checks pass, and reverse-image search finds nothing.
WHY THE USUAL CHECKS FAIL
EXIF is intact. The device is real. The location is real. Only a 300×200 pixel region of the image is a lie.
After a storm, claim volume spikes and processing thresholds relax to maintain service. This is when roof, water, and wind-damage photos flood in from claimants whose property was in the storm's path but undamaged.
THE OPERATIONAL TRAP
The surge that makes fast payment a service commitment is the surge that makes fabricated evidence profitable.
Proof of loss value is a document issue. Repair quotes and invoices are now generated end-to-end, ensuring consistency and accuracy.
WHAT SURVIVES A FORMAT CHECK
A regenerated invoice has no tampering to find. The line items add up. The vendor exists. The job never happened.
Fabricated medical invoices, treatment summaries and lab documentation submitted in support of health, life and disability claims — structurally valid, correctly coded, and describing care that was never delivered.
CROSS-REFERENCE
See also: Healthcare & Digital Health, where the same document class is attacked from the provider side.
RECEIPT DETECTION MODULE
The photo establishes that a loss occurred. The receipt establishes what it was worth — which is why inflating a claim is a document problem, not an image problem. Contents claims, repair reimbursements and proof-of-ownership documents all run through the same PDF detection engine.
See the Fake Receipt Detector$16.8M
in fraudulent receipts caught by one enterprise customer in a single 30-day period, surfaced by batch-scanning receipts that had already cleared human review.
ONE ENTERPRISE CUSTOMER · 30-DAY PERIOD · FIGURE PENDING INTERNAL VERIFICATION BEFORE PUBLISH
THE ENGINES
Photos and documents are different forensic problems, so we solve them with different models behind two dedicated APIs. You call the one that matches the artifact — and you review both in one dashboard, under one contract, through one vendor review.
IMAGE DETECTION API
Built for the failure mode that matters most in claims: not the fully generated image, but the authentic photo with fabricated damage added.
PDF DETECTION API
Forensics for the documents that set the claim's value: invoices, estimates, receipts, medical bills and ownership records.
WHERE IT SITS
The earliest useful place to score evidence is before a number is attached to it. Detection runs as a step in intake, not as an audit after the fact.
STEP 01
Claimant uploads photos and, for proof of loss, supporting documents.
STEP 02
Image API on the photos, PDF API on the documents. One to two seconds, end to end.
STEP 03
Straight-through processing for clean claims; SIU referral, with evidence, for flagged ones.
STEP 04
Heatmap, forensic report and model version attach to the claim record for appeal, litigation or regulator review.
DEFENSIBILITY
A confidence score is not a finding. What an insurer needs is an artifact that survives an appeal, a DOI complaint and, eventually, a courtroom.
Region-level heatmaps. The report shows where the image failed, which is the difference between an allegation and an exhibit.
Uniform thresholds across every claimant. The same model at the same threshold on every claim — the substance of a fair-claims-practice defense.
Retained model provenance. Version and timestamp are stored with the verdict, so a finding can be reproduced and explained years later.
Retroactive batch audit. Run detection across claims you already paid. This is where the number in the receipt module came from.
COMPLIANCE & DEPLOYMENT
SOC 2 Type II
Independently audited security, availability, and confidentiality controls, renewed annually.
ISO 27001
Certified information security management across the full platform and API surface.
GDPR
DPA available. Enterprise customers can pin processing to the EU.
Global Infrastructure
UK, EU and on-premise deployments for Enterprise.
TECHNOLOGY FEATURED IN
Get full forensic reports - heat maps, key indicators, and detailed descriptions.
Free
Try the full engine
Starter
$0.03 / result - $290/yr
For individuals and small teams
Professional
$0.02 / result - $990/yr
For teams in production
Business
$0.01 / result - $3,990/yr
For high-volume operations
No hard quota walls — usage is simply metered at your plan's rate. Custom contracts, SLAs, DPAs, and hosting environments are available exclusively on Enterprise. All plans include SOC 2 Type II compliance.
FAQ
Send us a sample from a closed period. We will run both engines across it and show you what cleared review — with region-level heatmaps on every finding.
NO CARD REQUIRED · UNLIMITED SEATS ON EVERY PLAN · SOC 2 TYPE II · ISO 27001 · GDPR
