AFCX Briefing

The Agentic Threat Landscape

Robert Miller · February 2026

6 months ago

Agents were demos

Today, they run 24/7, autonomously

100,000+

GitHub Stars

OpenClaw: Open source, self-hosted, you own your data

Fastest-growing agent framework in history

This is real

Meet George

My AI agent — running autonomously for weeks
Monitors markets, writes code, manages tasks
24/7 operation, no human in the loop
Built with the same tools available to anyone

The same tools I use
are the same tools
bad actors will use

The Threat Landscape

⚡ Scale

One person can run 1,000 scam agents simultaneously

🔄 Persistence

Never sleeps, never slips up, never gets tired

🎭 Sophistication

Voice cloning + agents = automated vishing at scale

🤖 Coordination

Agent-to-agent protocols enable swarms that communicate

30 minutes

Setup time

Free tools. No technical barrier.

A motivated scammer can deploy an agent swarm today.

How do you detect
an agent vs. a human
in your data today?

Let's discuss

What fraud patterns already feel AI-assisted?
Are you seeing volume spikes that suggest automation?
What does your threat model look like when every scammer has a swarm?
What's blocking on-prem AI adoption for AFCX?

Robert Miller

linkedin.com/in/rm16
@rob_miller87

George lives at agent-george.com