Apate.AI's bots have talked to scammers 2.5M times, extracted 250K intel leads
机器之心 · wechat · 2026-10-19
Australian company Apate.AI, spun out of Macquarie University security research, deploys AI agents posing as potential victims — answering scam calls, replying to phishing, joining investment groups — to extract intelligence from fraudsters.
Model and numbers
- 2.5 million conversations with scammers, wasting 17 years of their time and surfacing 250K+ intel leads (bank accounts, crypto wallets, phishing sites, phone numbers).
- Honeypot thinking at scale: agents mimic different ages, genders, accents; clients can task agents to obtain specific accounts or wallet addresses.
- Founder Dali Kaafar's inspiration: a 44-minute call he once wasted on a scammer — dragging out interactions raises scammers' cost (3-min vs 30-min exchanges mean a 10x difference in daily output).
Traction
- Customers include Commonwealth Bank of Australia and telecom TPG: TPG routed 590K suspicious calls; the bank reported a 76% drop in scam losses; one bank intervened in 22 scams via proactive agents.
- $8M+ seed round raised; US and Europe expansion planned.
Context: FBI IC3's 2025 report logged 22K+ AI-involved complaints with $893M losses. AI lowers costs for both attackers and defenders, but extracted intel still requires verification and institutional cooperation.
More from Safety
- GPT-4.1 with RAG hits 92.1% accuracy reviewing clinical trial informed consent forms — lamechthinkbig · 2026-10-19
- Claude's answer to Musk's Grok-DMs tip nails the prompt injection risk — alex_verem · 2026-10-19
- Musk warns many will be 'one-shotted' by super-persuasive superintelligences — elonmusk · 2026-10-19
- Gemini's critical summary of a safety report surfaces downplayed residual risks — davidmanheim · 2026-10-19
- Red teamers share AI safety findings that seemed harmless until they weren't — Plastic_Secret1303 · 2026-10-19
- Why aren't AI companies sued for unauthorized crawling and agent harm? — warfunder · 2026-10-19