Researcher warns self-replicating AI agent botnets are a near-term risk

joshua_saxe · x · 2026-09-15

joshuasaxe doubles down on his contested claim that an exponentially self-replicating agent swarm is very possible, with motivated actors today. Seed agents could opportunistically hack devices and steal API keys across 12+ inference providers to power the botnet, and hide inference within benign traffic on victim networks. Justin Halford, amplifying, says given Hugging Face and other recent incidents, this is a near-term risk.

Related event: Security researcher argues self-replicating AI botnets are feasible today(6 posts)→

Original post →

More from coding & agent

coding & agent channel →