AI agent swarms have 3x the viable attack paths of human hackers, Cogent warns

cneuralnetwork · x · 2026-10-08

Cogent launched Attack Path Analysis, deploying AI agents to find and fix machine-viable attack paths before adversaries exploit them.

Key points

In July, 700 AI agents breached Hugging Face's production infra, taking 17,600 actions, harvesting credentials, and moving laterally
Z.ai's open-weight GLM-5.3 reportedly builds working end-to-end exploits nearly as often as top models, with safeguards removable for a few thousand dollars of compute
Exploited vulnerabilities account for 31% of initial access in Verizon's 2026 DBIR
Cogent Research: every attack path viable for a human maps to three viable for AI agents, which never abandon long chains

The product targets chained routes—exposed edge device to stored credential to overprivileged role to crown jewels.

More from Companies & People