Programme PS-PAK-FRAUD-2026-001

An AI defence layer for Pakistan’s USD 9.3 billion annual fraud crisis.

Phoenix Sentinel is a sovereign, Pakistan-native fraud-intelligence platform built for banks, mobile wallets, regulators and citizens. We combine agentic AI, transaction-graph analytics and Urdu and Roman-Urdu phishing classifiers into a real-time defence layer that foreign vendors cannot replicate.

Operating EntityPhoenix Research and AI Systems (SMC-Private) Limited
SubmissionWorld Startup Championship 2026
CategoryScience, IT and Engineering
USD 9.3B
Annual losses to financial scams
Global Anti-Scam Report 2025
2.5%
Of Pakistan's GDP lost to scams
Feedzai · GASA
35%
Year-on-year surge in cybercrime 2025
Business Recorder
350
Cybercrime investigators for 160,000+ cases
NCCIA disclosure
The Crisis

Pakistan is losing more to digital fraud than it borrows from the IMF.

The numbers cross-validate across the State Bank, the National Cyber Crime Investigation Agency, the Banking Mohtasib, and independent international monitors. The crisis is structural, growing, and currently unaddressed by any Pakistan-native AI defence layer.

Magnitude

9.3B USD/year

Annual losses to financial scams — equivalent to 2.5% of Pakistan's GDP, and roughly one-third larger than the latest IMF programme tranche.

Velocity

150,542complaints / 2025

Cybercrime complaints filed with NCCIA in 2025 alone, of which 81,996 were financial fraud. Year-on-year surge of 35 percent.

Source: National Cyber Crime Investigation Agency, January 2026 NA disclosure
Capacity Gap

350investigators

Roughly 350 NCCIA cybercrime investigators handle more than 160,000 cases. Conventional enforcement cannot scale; AI-augmented defence is the only viable response.

Source: NCCIA / Banking Mohtasib Annual Reports 2024-2025
Regulatory Tailwind

SBPbiometric KYC mandate

In October 2025, the State Bank mandated biometric verification across every JazzCash, EasyPaisa, SadaPay and digital-bank account. The procurement window for fraud-defence platforms is now open.

Source: SBP Circular · ProPakistani, October 2025
Language Surface

7languages of attack

Urdu, Roman Urdu, English, Punjabi, Sindhi, Pashto and Saraiki phishing patterns dominate Pakistani scam traffic. Foreign vendors built for English-only surfaces miss the bulk of attacks.

Source: Phoenix Sentinel programme dataset PS-PAK-FRAUD-2026-001
Market White Space

0Pakistan-native AI defence

Across academic papers and industry registers, no Pakistan-headquartered AI fraud-detection platform of scale has been identified. The defensible position is wide open.

Source: Nature 2026 · Asian Bulletin of Big Data Management · Crunchbase
The Defence Layer

Three layers of defensibility no foreign vendor can replicate inside two years.

Phoenix Sentinel is not a thin wrapper over a foundation model. The defensibility sits in three layers that take years of Pakistan-resident engineering and regulatory relationships to assemble.

Layer 01

Pakistan-Native Models

Fine-tuned models trained on Pakistan-resident fraud data, evaluated against named benchmarks, with native classifiers for Urdu, Roman Urdu, Punjabi, Sindhi, Pashto and Saraiki phishing surfaces. Foreign English-only models miss the bulk of the attack distribution.

Layer 02

Sovereign Data Plane

A growing data asset of Pakistan-specific fraud patterns covering SIM-swap, OTP social engineering, Modaraba affinity fraud, Hajj operator fraud, Ponzi schemes, BEC plus deepfake, government impersonation, and property-society fraud. Twelve typologies, all locally validated.

Layer 03

Integration Depth

Native integrations with the Pakistani banking and mobile-wallet stack (HBL, UBL, MCB, Allied, Meezan, NBP, Bank Alfalah, Standard Chartered, Faysal, JazzCash, EasyPaisa, SadaPay, NayaPay, UPaisa) and direct triage routing to NCCIA, SECP, SBP, FIA and PTA.

Working Prototype · v0.1.0

Try four working modules of the Phoenix Sentinel platform.

The demos below run live in your browser against synthetic Pakistan-context data. The full prototype (FastAPI + Server-Sent Events + Operator Console) is deployed separately. All twelve scam typologies, ten cities and seven languages are modelled. No real PII is involved.

PS-CCOperator Command Centre
Live alert feed 0 alerts 0 high-risk
Conviction Rate
0.0%
Target: 12%
Recovery Rate
0.0%
Target: 45%
Mule Freeze
0s
Target: 60s
Awareness
0.0%
Target: 85%
5BD Refund
0.0%
Target: 90%
Deepfake KYC
0.0%
Target: 100%
Live Fraud Alerts0
PS-06Geolocation Fusion · Pakistan Heat Map
0 incidents0 cities
City Aggregates
Tap any marker for incident detail. CDR / IP / Wi-Fi fusion in production deployment.
PS-03Mule-Account Detection · Transaction Graph
0 accounts 0 edges 0 rings
High-risk Accounts0
Detected Rings0
Frozen (this session)0
Funds at risk (PKR)0
Synthetic data. Production deployment requires warrant-gated controls per PS-11.
PS-09Citizen Reporting Portal · Automated Triage
Try submitting a sample report

Report a scam

How auto-triage works

Phoenix Sentinel's triage engine assigns each public report a severity score based on the scam typology and the amount at risk, then routes it to the correct Pakistani agency:

  • SIM-swap / OTP → SBP · PTA · NCCIA
  • Ponzi / Modaraba / Crypto → SECP · NAB
  • BEC / Deepfake → NCCIA Cyber Cell
  • Fake recruitment → BEOE · FIA
  • Hajj / Umrah fraud → Ministry of Religious Affairs · FIA
  • Property → Provincial Authority (LDA/CDA/RDA/SBCA)
SLA: Critical 4h · High 24h · Standard 72h
The Platform

Twelve modules. Four shipped in the v0.1.0 prototype.

Phoenix Sentinel is architected as twelve interlocking modules across detection, prevention, recovery and governance. The 36-month programme sequences module delivery against named SBP, SECP, NCCIA and Banking Mohtasib milestones.

PS-CC
Operator Command Centre
Real-time alert feed, KPI dashboard, severity tracking, bulk-action interface for fraud-defence analysts.
Shipped v0.1.0
PS-01
Transaction Velocity Scoring
Sub-second risk scoring on every transaction across banks, wallets and Raast settlement.
Planned
PS-02
SIM-Swap Interception
PTA-integrated SIM-swap detection with telco-side carrier validation and pre-emptive freeze.
Planned
PS-03
Mule Network Detection
Transaction-graph analytics with ring discovery, node-level freeze and IBAN-cluster tracing.
Shipped v0.1.0
PS-04
Urdu and Roman-Urdu Phishing Classifier
Native NLP models for seven Pakistani languages; outperforms English-only foreign models.
Planned
PS-05
Deepfake KYC Defence
Liveness, voice-clone and synthetic-face detection layered onto bank and wallet KYC stacks.
Planned
PS-06
Geolocation Fusion
Pakistan-resident CDR, IP and Wi-Fi triangulation; suspect-cluster mapping by city and corridor.
Shipped v0.1.0
PS-07
Investment-Fraud Surveillance
Ponzi, Modaraba and crypto-platform surveillance with automatic SECP and NAB referral.
Planned
PS-08
BEC and Deepfake Defence
Business email compromise interception with executive-impersonation and deepfake voice flags.
Planned
PS-09
Public Reporting Portal
Citizen-facing intake in seven languages; automated triage and agency-routing to the right authority.
Shipped v0.1.0
PS-10
5-Banking-Day Refund Engine
Banking-Mohtasib-aligned refund automation with cross-bank settlement and audit trail.
Planned
PS-11
Warrant-Gated Governance
Audit, data-protection, lawful-intercept and warrant-gated governance controls per PDPB.
Planned
The Company

Built inside Phoenix Research and AI Systems.

Phoenix Sentinel is a programme of Phoenix Research and AI Systems (SMC-Private) Limited, a SECP-licensed Pakistan-resident research and intelligent-systems organisation incorporated under the Companies Act 2017.

Phoenix Research and AI Systems (SMC-Private) Limited

A frontier artificial-intelligence research and intelligent-systems organisation operating across South Asia, the Gulf Cooperation Council and East Africa. The studio architecture spans AI-as-a-Service (Phoenix International), healthcare and special-needs (Phoenix Minds), Mars-mission decision systems (MARS-MINDS) and now sovereign cybersecurity through Phoenix Sentinel.

IncorporationCompanies Act 2017
RegulatorSECP, Pakistan
HeadquartersKarachi, Sindh
Programme CodePS-PAK-FRAUD-2026-001
StageIdeation · Pre-revenue by design

Founder & Chief Executive

Ruksh E. Ibadat
Senior Artificial Intelligence Engineer · Founder

Founder of Phoenix Research and AI Systems and architect of the Phoenix Sentinel programme. Leads engineering, research and commercial direction across the studio's four ventures.

Built for Pakistan’s banks, wallets, regulators and citizens.

Phoenix Sentinel is currently competing in the World Startup Championship 2026 under the Science, IT and Engineering category. Design-partner conversations are open with banks, mobile-wallet operators, regulators and consumer-protection authorities.