finance
AI is Now the Biggest Threat to Your UPI Transactions — Here's What the Data Says
A
AITian EditorialMay 2, 2026
about 2 months ago
## Key Findings at a Glance
| Stat | Figure |
|------|--------|
| UPI fraud cases (2025) | ~28 lakh |
| Bank fraud losses FY25 | ₹36,014 crore |
| Indians hit by AI deepfake scams | 47% |
| Rise in deepfake fraud attempts | 2,137% |
---
## The Seven AI-Powered Threats
**1. Deepfake Voice & Video Fraud** — AI can clone a voice from just 3–10 seconds of audio. Scammers impersonate relatives in distress to trigger urgent UPI transfers. India is projected to lose ₹70,000 crore to deepfake fraud in 2025 alone.
**2. GenAI Phishing, Smishing & Vishing** — AI has eliminated grammar errors and telltale signs from phishing. 82% of phishing emails now contain AI-generated content; GenAI-driven phishing surged 1,265% in a single year.
**3. Synthetic Identity & Mule Account Networks** — AI fabricates fake Aadhaar/PAN cards and liveness-check videos. Over 1.1 million suspected mule accounts have been identified. "Golden Aadhaar Kits" sell on the dark web for ₹500–₹2,000.
**4. Adversarial Attacks on Fraud Detection AI** — Criminal networks use adversarial ML to probe bank fraud-detection systems, crafting transactions that appear normal until the final high-value fraudulent transfer is made — by which time the pattern has been whitelisted.
**5. Fraud-as-a-Service** — Voice cloning now costs $5 (down from $50,000 in 2023). AI scam toolkits, synthetic identity rentals, and mule networks operate via Telegram. AI scams surged 1,210% in 2025.
**6. Real-Time Fund Dispersal Outpacing Detection** — UPI processes transactions in 15 seconds. By the time a fraud alert triggers, funds have moved through 3–5 mule layers and are irrecoverable.
**7. Quantum Computing — The Long-Horizon Threat** — Nation-state adversaries are harvesting encrypted UPI data today to decrypt later using quantum computers. 70% of executives expect a quantum cyberattack within 5 years. NIST mandates deprecation of RSA/ECDSA by 2030.
---
## Policy Imperatives
- Mandatory AI deepfake detection in UPI apps
- Real-time cross-bank fraud data sharing
- Post-quantum cryptography migration across all 400+ NPCI participants
- Legal reform criminalising AI-generated synthetic identities
- Mass public awareness campaigns targeting rural and elderly users
*Study compiled: May 2, 2026 | Sources: NPCI, RBI, CERT-In, I4C, DST, Google GTIG, NIST, and 27 verified academic/investigative sources.*