AI Anti-Fraud Agents Engage Scammers in 2.5M Calls
Scammers dial relentlessly — AI answers back. 2.5 million conversations conducted. 250,000+ actionable fraud intelligence records extracted.
The Rising Toll of Telecom Fraud
How long has it been since you last answered an unknown call? For many, Apple’s call filtering feature is now permanently enabled.
While this blocks scam and sales calls, it also risks silencing legitimate ones — like package deliveries. Yet the trade-off feels necessary: fraudsters rotate numbers endlessly; we can’t mute our phones forever.
Enter a novel countermeasure — not human vigilance, but AI-led engagement.
Apate.AI: Deploying AI as Digital “Chaoyang Masses”
Sydney-based Apate.AI, founded in 2024 from cybersecurity research at Macquarie University, deploys AI agents that impersonate potential victims — answering scam calls, replying to phishing emails, clicking malicious links, and even infiltrating investment Telegram groups.
Their mission? Let the scammer speak — then extract intelligence.
Key Performance Metrics (as disclosed):
- ✅ 2.5 million scam interactions completed
- ⏱️ ~17 years of cumulative scammer time consumed
- 📊 250,000+ verified fraud intelligence items collected:
- Bank accounts & cryptocurrency wallet addresses
- Malicious domains & toll-free numbers
- Fraud scripts, impersonated brands, and social engineering tactics
How It Works: From Honey Pot to AI-Powered Threat Intelligence
Apate.AI reimagines the classic cybersecurity honeypot — but for fraud ecosystems.
| Traditional Honeypot | Apate.AI’s Approach |
|---|---|
| Simulated server to lure hackers | Simulated human personas to engage fraudsters |
| Monitors technical attack vectors | Captures linguistic, behavioral & operational fraud patterns |
| Requires manual threat analysis | Automates extraction, normalization & tagging of IOCs |
Core Capabilities:
- 🎭 Persona Simulation: Agents adapt age, gender, accent, language fluency, and emotional tone to mimic real targets.
- 📞 Multi-Channel Engagement: Voice calls, SMS, email, web forms, and social media DMs.
- 🧠 Goal-Directed Dialog: Predefined objectives (e.g., “Extract bank account number used in fake CBA support call”) guide conversational strategy.
- 📋 Intelligence Pipeline: Raw dialogue → structured IOC extraction → enrichment → analyst-ready report.
Real-World Impact: Banking, Telcos & Government Adoption
Apate.AI has moved beyond lab validation into production environments across Australia:
Verified Customer Outcomes:
- 🏦 Commonwealth Bank of Australia: Reported 76% reduction in fraud losses after integration.
- 📶 TPG Telecom: Routed ~590,000 suspicious calls to Apate.AI agents.
- 🏛️ Unidentified Australian bank: Proactively intervened in 22 active scams, preventing fund transfers.
These results reflect collaborative defense: Apate.AI delivers intelligence; banks and telcos execute takedowns, freezes, and alerts.
Strategic Advantage: Raising the Cost of Crime
Beyond intelligence, Apate.AI exploits economic friction in fraud operations:
| Avg. Interaction Time | Max. Daily Interactions (8 hrs) |
|---|---|
| 3 minutes | 160 |
| 30 minutes | 16 |
By extending average call duration 10×, AI agents degrade scammer ROI — turning high-volume, low-effort attacks into resource-intensive dead ends.
Crucially, each interaction serves dual purposes:
1. 🕒 Wastes attacker time (raising operational cost)
2. 📥 Harvests forensic data (enabling proactive response)
Challenges & Limitations
Despite promise, scalability brings complexity:
⚠️ Data ≠ Intelligence: Volume requires rigorous filtering to discard decoys, false leads, and obfuscated info.
⚠️ Actionability Gap: Extracted wallet addresses or domains only help if banks, exchanges, and law enforcement act swiftly.
⚠️ Adversarial Adaptation: As scammers deploy LLM-generated scripts or deepfake voices (FBI IC3: 22,000+ AI-related complaints in 2025, $893M losses), defenders must continuously evolve.
Apate.AI trains models on proprietary fraud dialog datasets and academic research — building a self-reinforcing feedback loop between field data and model improvement.
The Future: Global Expansion & Systemic Integration
With $8M seed funding, Apate.AI is expanding into the US and EU — tailoring compliance, languages, and regional fraud typologies.
Its vision extends beyond point solutions: A unified anti-fraud layer, integrating with SIEMs, banking core systems, and national cybercrime units.
As AI reshapes both offense and defense, Apate.AI exemplifies a critical paradigm shift —
Don’t just detect fraud. Engage, observe, extract, and disrupt — before the first victim clicks.
Source: Machine Heart (via Sia)