A phone alone doesn't prove an accident. Merklit proves individual facts about the claimant, vehicle, place and time, then combines independent facts to support, contradict or qualify the incident narrative. Weak evidence routes to human review, never automatic rejection.
[01]
Reduce ambiguity
Separate verified observations from claimant statements and unavailable evidence.
[02]
Strengthen provenance
Bind fresh media, time, location, actor and device evidence into one tamper-evident record.
[03]
Support review
Route contradictions and weak evidence to human review instead of automatic claim decisions.
02 / proof of the problem
This is not a hypothetical risk.
Recent Indian cases show alleged manipulation at multiple points in the claim chain: identity, records, vehicles, drivers, photographs and settlement workflows.
01Sambhal network
02Police records
03MACT claims
04Damage photos
NDTV / PTI · 05 JUN 2025 · FOLLOW-UP 13 JAN 2026
₹100+ cr
Estimated alleged insurance fraud reported across 12 states
Investigators said deaths were passed off as road accidents to trigger insurance payouts.
The Sambhal-linked investigation described forged or altered records, policies created using vulnerable or deceased people, and staged accident narratives. Officials reported 17 FIRs, including four for murder. By January 2026, follow-up reporting described 70 arrests and ₹12 crore in attached assets.
Accident records allegedly changed before settlement.
An SIT found suspected irregularities in 13 fatal road-accident cases. Its findings described insured vehicles being substituted for the vehicles involved and driver details being changed in investigation records.
Insured vehicles allegedly planted in fake MACT claims.
The Supreme Court raised concern over an alleged wider racket involving insured vehicles being strategically shown in accident claims. The report describes suspected coordination among multiple stakeholders to fabricate compensation claims.
₹4.30 crore alleged fraud using reused damage photographs.
A CBI case based on an insurer complaint alleges fraudulent settlement of 411 claims over more than three years. The complaint describes reused photographs and suspected coordination involving vehicle owners, garages, surveyors and others.
Source note: these reports concern allegations, investigations, suspensions or court proceedings. They do not by themselves establish the guilt of any person named.
Pattern detectedManipulation can cross people, vehicles, records and media.Next: proof by assertion ↓
03 / proof model · scroll to step through
Five assertions. No black-box score.
Each fact keeps its own evidence basis and verdict, so reviewers can see exactly what is supported, and what is not.
{{ counter }}assertion progress
Assertion indexstate
assertion record activereported independently · human review preserved
default_response > insurers learn whether the claimed region and evidence requirements were supported, never an unrestricted dump of personal location history.
Pipeline output → deployment input
One evidence packet. Six ways to deploy it.
Claim configured
Channel ready
The assurance model stays consistent whether capture starts from a link, an SDK, a field app or a telematics source.
05 / deployment
One platform. Multiple assurance levels.
01 / 06
assurance surface
NO_INSTALLD-01
[↗]
ClaimLink
SMS/WhatsApp session with live camera, foreground location and immediate hashing.
HOSTEDD-02
[▣]
Branded claim site
A managed claim journey with guided capture, document upload and full insurer branding.
EMBEDD-03
</>
Web SDK
Embeddable components and configurable capture steps for existing portals.
OEM, fleet and OBD connectors, plus police, garage, towing and dashcam import.
06
06 / integrity and legal boundary
Tamper evidence, correctly framed.
The integrity layer proves what was committed, when it existed and whether it changed. It does not turn a device, sensor or blockchain into a truth oracle.
[ Not a truth oracle ] Evidence quality remains explicit
Recommended integrity sequence
06 controlled steps
01Hash at captureBind components to the fresh claim session.
02Sign manifestHardware-backed where supported, service key otherwise.
03Create claim rootCombine evidence-item roots into one commitment.
04Licensed-CA timeObtain an Indian trusted timestamp.
05Publish batch rootSend only the tenant-safe batch root to Base.
06Preserve verifierKeep receipt, inclusion path, versions and status.
The proof establishes
The exact evidence commitment existed by a recorded block.
A later change would produce a different hash.
The public anchor can be checked independently of Merklit.
The proof never establishes
That GPS or another sensor input was truthful.
That the claimant personally held the device.
That an accident occurred merely because a hash exists.
[ Legal boundary ]
The export is a court-supporting electronic evidence package, not guaranteed admissibility. Admissibility and evidentiary weight remain subject to proper procedure, case facts and legal review.
Encryption
AES-256-GCM per evidence object, tenant-specific KMS keys and optional BYOK.
On-chain policy
PII, raw media, policy numbers and coordinates never enter the blockchain.
Disclosure
Regional results by default. Exact location and full reports require an audited escalation.
Legal package
Manifest, signatures, licensed-CA time, inclusion proof, audit trail and a configurable Section 63 workflow.
07 / measurable pilot
Prove the value before changing a claim decision.
Run a shadow pilot on 500–2,000 motor claims without changing approval decisions. Compare cycle time, recapture, contradiction yield and review effort against your current process.
