Login
Fraud prevention

The real-time fraud prevention layer.

Identity, device, telecom, behavior and watchlists in a single decision, for onboarding, Pix, registration, claims and crypto.

See use cases
AI agents

The AI that investigates, decides and explains.

Four specialist agents work alongside the prevention desk: read documents, map networks, classify news and generate ready reports. Every decision comes with evidence and a trail.

Live · 138 events today

Risk Hub

247monitored
12critical
36elevated
Agent · Anti-deepfake

Blocks synthetic faces and injection.

Active and passive liveness, document anti-tampering and AI-generated image detection. Proprietary models vote alongside Unico, Caf and Facephi for evidence.

LivenessAnti-injectionDeepfakeOCR doc
< 400msp50 per decision
Agent · Fraud network

Exposes money mule accounts and rings.

Combines device fingerprint, address, beneficiary network and behavior in a single graph. When one account burns, linked ones show in the same search.

DeviceGraphMule accountUBO
9→4accounts in cluster · 4 confirmed
Agent · Fraud report

Ready report arrives at the desk.

Detects tampered documents (DMHO, ID, vehicle title), generates fraud hypothesis with consolidated evidence and auditable dossier. The desk reviews exceptions, not a stack of PDFs.

DMHOAI reportHypothesisTrail
96.4%accuracy in DMHO extraction

+ 16 agents ready in the library · or build your own in the agent studio

Five use cases

One workflow investigates
5 fraud types.

Onboarding, Pix, money mule, claims and AML crypto. The engineering is the same. What changes is the policy, the signals and the bureaus.

01 / 05Onboarding & Identity
01 · ONBOARDING

Onboarding & Identity

Facial biometrics, liveness, document OCR and deepfake detection in a single decision. True identity in seconds, no friction for the legitimate customer.

Cross signals
  • Biometrics + liveness
  • Deepfake detection
  • Document + watchlists
Regulations covered
Res. CMN 4.893Res. CMN 4.753Res. CVM 50COAF · FilingLGPD
Workflow builder

You compose the steps like a flow.

Each signal becomes a step that writes to the same dossier. Chain identity, device, network, watchlists and score in your policy order and branch the outputs. No code.

Differentials

Why VAAS, not another anti-fraud engine.

Not a bureau, not a static rule, not a POC. It is prevention infrastructure in production, with an auditable trail and a model calibrated for Brazil.

Device + behavior

Money mule accounts, mules and rings caught by device and behavior, device fingerprint, repeated address, shared beneficiary.

EVENT · PIX
Accounts on device9
Mules confirmed4
device · ZOOX

Signals calibrated for Brazil

Pix, money mule accounts, SIM swap, OFAC, adverse media. Not an imported rule set, Brazilian data and models.

  • Pix · DICTbeneficiary · velocity
  • Telecom BRSIM swap · ownership
  • Official biometricsRFB · liveness · deepfake
  • COAF · OFAC · PEPwatchlists · adverse media
38+
bureaus integrated, normalized into typed variables for the decision engine

In production in a regulated environment

BACEN bank, multi-vertical fintech, marketplace, acquirer and SUSEP insurer. Real volume, sensitive data, end-to-end auditable trail, ready for BACEN, SUSEP and external audit.

Bank · Pix at scale
Onboarding + transactional · BACEN
live
Multi-vertical fintech
Identity · credit · COAF
live
SUSEP insurer
Claims · DMHO · registration
live
Ready?

Decide in seconds.
Start with a meeting.

In 15 minutes we show how VAAS works in your scenario, with your rules, your data, your volume.

Explore the platform