AI Bots Turn the Tables on Phone Scammers at Scale

An Australian company named Apate — after the Greek goddess of deception — is flipping the script on phone fraud by deploying hundreds of thousands of AI-driven personas that answer scam calls, infiltrate messaging groups, and reply to malicious texts. The bots are engineered to sound like plausible victims: they hesitate, ask questions, sometimes claim to have WhatsApp, sometimes hang up and promise to call back. The goal is simple — keep the criminal on the line as long as possible while quietly logging every URL, money-mule account, and bank detail the scammer reveals.

Founder Dali Kaafar says the platform now runs roughly 350,000 concurrent bots and has amassed more than 250,000 actionable intelligence artifacts in real time. Banks and telecom carriers integrate the service to divert suspicious traffic away from real customers. In controlled tests, the bots sustained conversations for over two hours, frustrating even seasoned red-teamers who tried to break character.

Why it matters for GPU / AI infrastructure

Running hundreds of thousands of low-latency, multilingual voice and text agents demands massive parallel inference capacity — exactly the workload profile that modern GPU clouds are built for. As anti-fraud AI moves from research demos to production scale, the need for elastic, cost-efficient GPU clusters becomes a strategic differentiator for security vendors and the carriers that host them.