Secure Talk: Using AI to Protect Global Communications & Privacy

Contents

Executive Summary

This summit brought together telecom leaders, fintech executives, and banking officials to address the trillion-dollar global scam epidemic through AI-driven solutions. The central narrative: scams and fraud have evolved from isolated incidents into industrial-scale, AI-powered threats requiring coordinated ecosystem-wide defenses—not isolated institutional responses. The event showcased Tanla Platforms' Wisely.ai platform as a concrete implementation achieving real impact at scale across Indonesia (IndoSat), India (BSNL), and the banking sector.

Key Takeaways

Scams are now industrial, cross-border, and require industrial-scale coordinated response. Individual AI models within institutions are necessary but insufficient; real-time data sharing across the ecosystem is the next frontier.
Behavioral/intent detection is the unsolved problem. Current AI excels at anomaly detection but cannot reliably stop social engineering. Until models can detect malicious intent earlier in the journey (before the transaction), scammers will adapt faster than defenses.
Mule accounts and organized money laundering rings are the lever that scales fraud. Stopping individual scammers is ineffective; the system must make it costly and risky to rent accounts or move stolen funds. This requires coordinated law enforcement + intelligence + financial controls.
The defense can never be fully unconstrained. Defensive AI must respect privacy, regulation, and customer experience; offensive AI doesn't. This structural asymmetry means the goal is not "stop all fraud" but "raise the cost and risk to scammers faster than they can adapt."
National-level coordination frameworks (like RBI's Digital Payment Intelligence Platform) are necessary but insufficient without real-time, cross-border intelligence sharing and enforcement. Scammers already operate internationally; defenses must too.

Key Topics Covered

Global Fraud Crisis at Scale: $1 trillion+ in consumer losses globally; 65% of Indonesians face spam/scam weekly; India's Supreme Court recently identified ₹54-56,000 crores in losses
AI as Both Weapon and Defense: Scammers use AI-powered voice cloning, synthetic identities, and automated phishing; defenders must use AI to counter, but operate under privacy/regulatory constraints that offenders don't
IndoSat's Transformation: CEO Vikram Singha's case study on converting fraud/spam from customer complaint to board-level strategic priority
Banking & Fintech Perspectives: Transaction monitoring, rule engines, and customer behavior profiling (ARPU, churn reduction); challenges of preventing mule accounts and account rentals
Ecosystem Fragmentation: Each institution sees pieces of the fraud journey (telecom sees initial signals; banks see transaction risk; e-commerce sees behavior patterns) but no single entity sees the whole picture
Regulatory & Law Enforcement Gaps: Data silos, insufficient inter-agency coordination, and inadequate enforcement despite known scam origins
BSNL's Parallel Initiatives: Spam/phishing blocking on SMS channel; AI-driven network intelligence; rural edge deployments with federated learning
Customer Experience vs. Security Trade-offs: Balancing legitimate friction (2FA, transaction alerts) with customer convenience and growth metrics
Data Sharing & National Coordination: RBI's Digital Payment Intelligence Platform (Mule Hunter initiative) as emerging national-level solution
Privacy-Preserving AI: Federated learning, SLMs (Small Language Models) at the edge, local data retention with distributed learning

Key Points & Insights

The Fraud Scale is Asymmetric: While 65% of Indonesians face weekly spam/scam attempts, actual prevention success rates remain low because scammers have adapted to industrial scale using AI, while defenses remain fragmented by institution and geography.
AI Offense Outpaces AI Defense: Offensive AI operates unconstrained; defensive AI is constrained by privacy, regulation, and customer experience requirements. This structural imbalance means scammers stay ahead until foundational data sharing and real-time coordinated intelligence exist.
Behavioral Intent > Anomaly Detection: Current AI models excel at detecting transaction anomalies (unusual amounts, off-pattern purchases) but cannot reliably detect malicious intent before social engineering works. Voice cloning, deepfakes, and personalized targeting exploit psychology faster than models adapt.
No Institution Controls the Full Journey: Telecom operators see initial signals (SMS/OTP spam); e-commerce platforms see browsing and purchase patterns; banks see transaction risk and account abuse; but none see the complete social engineering → payment → money laundering chain. Data silos are the root enabler of fraud.
Mule Accounts & Account Rentals are the Scaling Problem: Beyond defrauding individuals, organized rings open multiple bank accounts (India Stack's easy KYC enables this) and rent them to fraudsters for ₹50,000–₹1.5 lakhs/month. This turns millions of willing low-income participants into unwitting accomplices, enabling industrial-scale money laundering.
Validated Business Case: IndoSat achieved 6-8 month ROI: ARPU +9% (vs. industry 3%), churn reduction from 3.6-3.7% to 1.6% for 90+ day cohort; estimated $500M in losses prevented within 6 months of Wisely.ai deployment.
Regulatory & Enforcement Gap Exists Separately from Technology Gap: Police and agencies know fraud origins but lack coordination/resources for prosecution. Technology cannot substitute for law enforcement will—"until there is fear of law," fraud persists.
Ecosystem Responsibility is Distributed but Accountability is Unclear: Banks regulate transactions; telecom filters messages; government oversees licensing; police enforce law; consumers must be educated. Responsibility is shared but no single actor owns the outcome, creating accountability diffusion.
Real-Time Coordinated Intelligence > Better Individual Models: The next frontier is not smarter AI models within institutions, but real-time data sharing across telecom, banking, payments, e-commerce, and law enforcement to detect patterns at national/international scale.
Privacy-Preserving Compute is Essential: Federated learning (data stays local, intelligence is distributed) and edge deployment of small language models allow pattern detection without centralizing sensitive financial/behavioral data—critical for public trust in national-level solutions.

Notable Quotes or Statements

Vikram Singha (IndoSat CEO) on recognizing the crisis:

"In 2024 itself, 5 billion USD worth of losses with Indonesians have been lost... 65% of Indonesians are facing spam or scam on a weekly basis. As a CEO, our role is not only to connect, our role is also to protect my 100 million customers."

Vikram Singha on the partnership philosophy:

"We don't need a vendor, we need a partner who can work with us and use AI to solve this real problem."

Nha (Amazon Pay India) on the fundamental mismatch:

"AI is helping detect anomalies. It is not helping detect the mal intent or the behavior. And unless until we solve the mal intent or the behavior, the scams will keep going."

Pratan (Bandhan Bank) on the systemic incentive problem:

"A fraudster asking my employee: 'Can you tell me how much you make a month? I'll give you double. Just share more data with me. I'll pay you ₹1,50,000 per account.' This is the problem we're facing."

Vipin (Moiqu CEO) on the need for national coordination:

"One entity cannot control scams. 99% of whatever scams our customers complain of are not the money stolen from Moiqu but from some other bank and come into Moiqu. Until data sharing happens at an India-level scale, you cannot identify patterns."

A. Robert Ravi (BSNL CMD) on the vision:

"Unless we have built a system where we confidently say to our citizens that you are 100% safe in my network, our job is not done. This is possible only when we bring in technology and play it across a platform which really intelligently builds this network."

Sanjay Kapoor (Tanla Board Member, ex-Bharti CEO) closing on trust as infrastructure:

"When over 100 million subscribers are protected by AI, when billions of communications are analyzed in real time, and when millions of malicious actors are stopped within the ecosystem, trust is no longer a promise. It becomes an infrastructure."

Speakers & Organizations Mentioned

Primary Speakers

Vish Gurmuk D. – Tanla Platforms (host/opening remarks)
Sanjay Kapoor – Board Member, Tanla Platforms; Ex-CEO, Bharti Airtel; Senior Role, GSMA
Vikram Singha – President, Director & CEO, IndoSat Orido Hutchison
Anuman Kar – Chief Customer Success Officer, Tanla Platforms (panel moderator)
Pratan – Executive Director & COO, Bandhan Bank
Nha Gi Mahatme – Director, Amazon Pay India
Vipin Pri Singh – Founder & CEO, Moiqu (digital wallet, 180M users)
A. Robert Ravi – Chairman & Managing Director, BSNL

Organizations & Institutions

Tanla Platforms – Core platform provider (Wisely.ai agentic AI system)
Carics (group company) – Anti-fishing tool, SMS scanning
Value First (group company) – Group entity
IndoSat Orido Hutchison – Indonesian telco, 100M+ customers
BSNL – Bharat Sanchar Nigam Limited (India's state-owned telecom)
Bandhan Bank – Private bank, 32M customers, 35 states
Moiqu – Fintech, largest digital wallet in India, 180M users
Amazon Pay India – Payments platform
Mastercard (via advisory board) – Referenced for scam data
RBI (Reserve Bank of India) – Regulatory body; developing Digital Payment Intelligence Platform (Mule Hunter initiative)
GSMA – Global mobile industry association
Bharti Airtel – Major Indian telco (context of Sanjay Kapoor's tenure)
HDFC Bank, Axis Bank, ICICI – Banks mentioned in context examples

Technical Concepts & Resources

Platforms & Systems

Wisely.ai – Tanla's agentic AI platform for identifying, preventing, and eliminating spam/scam; live at IndoSat, BSNL, and Indian banks
Anti-Fishing Tool (Carics) – Scans SMS origins, analyzes links, identifies fake vs. genuine content using AI algorithms
Awani System – BSNL's one-touch voice agent for customer interaction
BSNL Recharge Expert System – AI-driven system for recharge automation
Digital Payment Intelligence Platform (Mule Hunter) – RBI initiative for cross-ecosystem data sharing

AI & ML Techniques

Anomaly Detection – Rule engines + transaction monitoring to identify out-of-routine transactions
Behavioral Profiling – Customer pattern analysis (ATM usage, transaction frequency, amount patterns)
Rule Engines – Sophisticated algorithmic frameworks for real-time transaction filtering
Voice Cloning – Scammer technology; clones voice for social engineering
Synthetic Identities – AI-generated fake profiles used in account opening fraud
Federated Learning – Data remains local; distributed learning without centralizing sensitive data (BSNL's future direction)
Small Language Models (SLMs) – Lighter, locally deployable models for edge computing (vs. large LLMs requiring cloud compute)
Deep Fakes – AI-generated synthetic media for fraud

Metrics & Measurements

ARPU (Average Revenue Per User) – IndoSat grew 9% vs. industry 3% within 6-8 months
Churn Rate – IndoSat reduced 90-day churn from 3.6-3.7% to 1.6%
Estimated Fraud Prevention – $500M in losses prevented within 6 months (IndoSat)
Spam/Scam Calls Stalled – BSNL: ~280 million blocked connections
SMS Volume in India – 65 billion SMS/month; 15 billion OTT messages/month

Infrastructure

GPU Clusters – IndoSat deployed proprietary clusters (GB200, H100) for local training and inference
Edge Data Centers – BSNL's BharatNet rollout enabling edge deployment of intelligence
Network RAN Intelligence – BSNL's AI-driven RAN (Radio Access Network) for pattern detection and customer-specific service allocation

Regulatory & Policy References

Sanchaya App – Government of India's SMS complaint platform
Digital Payment Intelligence Platform (RBI) – Data-sharing initiative across financial ecosystem
GSMA Guidelines – Mobile industry standards and policy frameworks
KYC (Know Your Customer) – India Stack's simplified digital KYC enabling rapid account opening (simultaneously enabling fraud at scale)
2FA/2FA Standards – Second-factor authentication inconsistency across institutions identified as gap

Data & Statistics Cited

Global Scam Losses: $1 trillion+ (consumers globally)
India: ₹54-56,000 crores (~$6.5-7B USD) identified by Supreme Court
Indonesia: $5 billion USD in losses in 2024; 65% weekly scam exposure
Digital Economy Projections: India $1 trillion by 2030; Indonesia >$100B GMV already
Digital Population: 5 billion people online globally; 2 billion coming online in South/Southeast Asia
Digital Payments: Expected to exceed $14 trillion annually by 2027
India SMS/OTT: 70% of scams originate from SMS channel; 65 billion SMS/month; 15 billion OTT messages/month

Structural Insights & Implications

Problem Architecture

Multi-Stage Attack Chain: Social engineering → Account compromise → Payment/transfer → Money laundering (mule accounts) → Cash out
Current Defense Fragmentation: Telecom (message filtering) ≠ Bank (transaction monitoring) ≠ E-commerce (behavior signals) ≠ Law enforcement (prosecution)
Data Silo Effect: No single entity sees the full journey; patterns visible at ecosystem level invisible at institutional level