AI news · October 15, 2026
Apate uses undercover agents to map scam operations
Fraud defence is shifting from classifying one message to collecting long conversations and campaign structure, with agents acting as controlled decoys.
- conversation claim
- 2.5M
- scammer time claim
- 17 years
- seed funding
- $8.15M
TechCrunch reported on October 14 that Apate uses thousands of AI agents as controlled scam victims across calls, messages, email and the web. The company says the agents have consumed millions of scammer messages and can help banks, telecom companies and governments see campaign structure rather than isolated indicators. Customer examples in the report include a telecom operator diverting scam calls and a bank reporting a large fall in losses, based on company and customer statements.
The defensible product pattern is the long interaction. A classifier can flag a message, but a decoy agent can reveal the next domain, script, payment request and handoff. This is a high-stakes system, so teams need strict scope, rate limits, evidence capture and a human review path. Treat the reported outcomes as customer claims until independently measured.
What you can do with it
If you work on fraud operations, start with a sandboxed decoy for one campaign type. Define the allowed channels, stop conditions, evidence fields and escalation path. Measure intelligence gained per analyst hour, not only the number of conversations held.
Our take
Apate makes an agent useful as an intelligence collector, not as a pretend customer for its own sake. The value is campaign visibility and analyst leverage. Governance and containment are part of the product because an uncontrolled decoy can create new operational risk.
Source: TechCrunch / Apate — Made With Models writes the brief; the reporting is theirs.