Riskernel - Fraud Surface OS | Investor Overview

Fraud detection is stuck in 2015. It asks "Is this transaction suspicious?" and stops there. Cybersecurity moved to attack surface management, deception, identity governance, and provenance years ago. We're importing those paradigms into fraud -- building the first platform that maps the fraud surface, plants deception signals, governs non-human economic identities, and executes auditable automated response. The scoring API is the wedge. The platform is the destination.

The Paradigm Shift: Event-Centric to Surface-Centric Fraud

Today's fraud tools score individual transactions. They wait for suspicious activity, then react. No visibility into what's exposed. No deception. No governance of AI agents entering commerce. No provenance on decisions. Response is either manual or uncontrolled AI.

Fraud Surface OS starts before the transaction. Map your fraud surface. Plant deception signals. Govern every economic identity. Validate customer environments. Execute deterministic, auditable response. Prove every decision to regulators. Simulate changes before production.

The Platform: 10 Engines, One Graph

Fraud Exposure Graph

Live map of exploitable surfaces -- domains, payout flows, refund loops, onboarding gaps, mule corridors. Scored by blast radius.

Fraud Deception Mesh

Honey IBANs, trap promo codes, decoy merchants, honey wallets. Near-zero-noise early warning of attacker intent and TTPs.

Economic Identity Governance

Identity, permissions, spend limits, and retirement for every non-human actor: AI agents, seller bots, refund bots, treasury scripts.

Attested Customer Runtime

Hardware-backed verification for high-risk ops. Detect overlays, remote control, clipboard abuse, automation.

Verifiable Interaction Authenticity

Signed capture for KYC, voice approval, consent. Provenance over detection -- prove where it came from.

Decision Provenance Ledger

Supply-chain-grade evidence per decision: signals, model version, policy, overrides, customer communication.

Fraud Remediation Language

Deterministic action vocabulary: quarantine, step-up, hold, callback. AI recommends; execution is bounded and auditable.

FraudSecOps Design Reviews

AI-powered abuse case generation for every new flow, product, and vendor integration before launch.

Need-to-Know AI Layer

Role-based visibility for fraud copilots. Prevents playbook, threshold, and PII leakage through AI itself.

FraudOps Digital Twin

Simulate threshold changes, vendor outages, scam waves, and policy shifts before production. Predict impact on losses, queues, and friction.

Why now: Authorized scams (pig-butchering, social engineering) are shifting from operational problems to liability and regulatory scrutiny. APP fraud reimbursement mandates are expanding globally. AI agents entering commerce create non-human economic actors nobody governs. Deepfakes make detection insufficient -- provenance becomes necessary. Regulators demand explainability current systems can't provide.

Phased Execution

Now

Scoring API wedge -- 47ms, explainable, single API call. Free shadow tests. 190+ leads in pipeline. Landing first customers.

Phase 2

Decision Provenance + Remediation Language -- Extend API with evidence chains and deterministic actions. Differentiate from commodity scoring.

Phase 3

Fraud Exposure Graph -- Analytics layer mapping each customer's fraud surface from scored transaction data.

Phase 4

Deception Mesh + Design Reviews -- Proactive fraud prevention. Honey signals + shift-left abuse analysis.

Phase 5

Full Platform -- Identity governance, attested runtime, AI governance, digital twin. Fraud Surface OS.

ICP & Moat

Initial ICP: Payment processors, BaaS platforms, digital banks, BNPL/lending fintechs, push-payment platforms. 50-500 employees, US-based. Processing financial transactions where fraud liability is direct.

Expand to: Marketplaces (payout + seller fraud), crypto platforms (AML + wallet fraud), insurance (claims fraud), cross-border payments.

Moat: The unified graph connecting exposure + deception + identities + runtime trust + remediation + provenance. Each engine feeds the others. Compounding defensibility that single-point fraud tools can't replicate.

Founder-market fit: 12 patents across fraud detection + cybersecurity + AI. The only founder profile that spans all three domains needed to build this platform.