Fraud management · PILOT READY

Stop the fraud. Keep the customer. Explain both.

Use Nyx to approve, review or decline any payment, login or account change before the money moves, with the reason attached. Built and ready to pilot on your data.

Run a fraud pilot See the engine underneath
NyxFraud decisions
1The request
Asked by your payment switch
INR 49,500.00
Rail
UPI, push
Payee
First seen
Account
Many new senders
2The decision PILOT READY
Review

New payee and mule-pattern inflow

  • Payee ruleupi_new_payee_velocityTriggered
  • ModelmuleCBM@v30.88
3The evidence record IN BUILD
Actor
Person · the account holder
Principal
Self · no delegation
Shown
Scam warning before paying
Decision
Review, then warn
What a dispute or an examiner reads
// POST /api/interface/decision/decide · 84ms†
{
  "action": "review",
  "reason": "new payee + mule-pattern inflow",
  "rule": "upi_new_payee_velocity",
  "model": "muleCBM@v3",
  "score": 0.88,
  "severity_band": "high",
  "audit_ref": "dec_EXAMPLE…",
  "latency_ms": 84
}
84ms† · audit_ref dec_EXAMPLE…REPRESENTATIVE · SAMPLE DATA · LATENCY TARGET <100MS†
Built for the authorization path
Card, UPI and account railsDevice, velocity, geography and behavior; consortium scores such as DPIP through an adapter.
Reasons with every verdictThe score, the rules that fired, the model version and an audit reference.
Replay before you decideYour own history, replayed where it lives, with nothing leaving your perimeter.
What the pack decides

Six kinds of fraud, one engine.

Mandate-scope checks on UPI Reserve Pay, e-mandates and agentic tokens are the agent-trust pack, built with the first design partner. The agent-trust pack →

Cross-border CNP authenticationRisk-based authentication under the RBI Authentication Directions 2025, from 1 Oct 2026.
Account takeoverLogins and account changes that do not fit the customer.
Mule detectionAccounts that receive fraud proceeds and pass them on.
Onboarding fraudApplications built on synthetic or borrowed identities.
Scoring before completionA verdict before the transaction completes, not after.
Scams and social engineeringPayments a customer is manipulated into making.
Two paths

The verdict stays lean. The investigation runs beside it.

01PILOT READY
Real-time decision path

A verdict inside the authorization path.

A hard timeout budget and a fallback verdict when it is breached. Fail-open or fail-closed is your policy, set in configuration.

02IN BUILD
Async investigation path

The case is assembled before an analyst opens it.

Related cases, device history, dispute context and policy evidence, gathered by a triage agent that recommends and never decides.

<100ms†scoring latency, in the authorization pathDesign target
30–50%†fewer false positivesDesign target
45 sec†investigations, vs minutes of manual review todayDesign target
3–5x†analyst throughputDesign target

† Design targets to validate in the first pilot, not yet independently benchmarked. When we have pilot numbers, they replace these.

A first pilot

Shadow first, enforce second, on your own history.

Compare the deployment models →
  1. ReplayYour own decision history, replayed where it lives, with nothing leaving your perimeter.
    false declines you could recovermule accounts you missedagent-originated traffic sharedisputes you could have defended
  2. ShadowNyx scores beside your existing checks, acting on nothing, so you can compare its calls with your team’s.
  3. EnforceNarrowly, on a scoped segment, with your fallback verdict and your rollback.

Deployment is your choice in every case: managed by us, in your cloud, or in your data center.

Questions

Before a first pilot.

What the pack decides, and what is still a target.

Is Nyx a fraud detection system?

Fraud detection and prevention is one of the decisions Nyx makes, and the one ready today. The fraud management pack scores each payment or account event in real time, before the money moves, and returns approve, review or decline with the reason attached. The same engine also decides agent payments through the agent trust pack, which is in build, and credit, disputes and AML, which are partner-led: engine primitives in place, each pack scoped with the first partner who needs it. Nyx is pre-launch: the fraud pack is built and ready to pilot, and its performance figures are design targets until a pilot measures them.

What does the pack decide today?

Card-not-present risk-based authentication, account takeover, mule detection, onboarding fraud, transaction scoring before completion, and scam and social-engineering patterns, across card, UPI and account rails.

What about the RBI deadline?

RBI's Authentication Directions 2025 require risk-based authentication on cross-border card-not-present transactions from 1 October 2026. That is a decision in the authorization path with a reason attached to it, which is what this pack does.

How do we try it without touching production?

With a shadow replay of your own decision history, run in the deployment model you choose with nothing leaving your perimeter. It produces four numbers: false declines recoverable, mule accounts missed, agent-originated traffic share, and dispute cases defensible with evidence.

Are the numbers benchmarked?

No. Scoring under 100ms, 30 to 50 percent fewer false positives, 45-second investigations and 3 to 5 times analyst throughput are design targets to validate in a first pilot. When pilot numbers exist, they replace the targets.

What comes back on every call?

The action, the score, the severity band, the policies and rules that fired, the model version and an audit reference. The canonical evidence record the reference points to is in build.

Fraud pack · pilot ready

Replay your history. See what Nyx would have decided.

Bring a slice of your decision history and your current thresholds. We will show you what Nyx would have decided, where it disagrees with you, and the four numbers you do not have today.

Run a replay Other decision domains

HELLO@NAXVERSE.AI · FOUNDER-LED, PROCUREMENT-READY · DUE-DILIGENCE PACK ON REQUEST