Intelligence about AI agents,
code security, and the gaps
tools miss.

Practical research for security engineers and developers building or reviewing AI-powered systems.

Threat Research
FeaturedDraft — Pending Review

Your SAST Passed. Your Agent Is Still Dangerous.

Traditional AppSec tools miss the agent security surface. Here's the gap and how to think about closing it before you ship.

SAST scanners look at the wrong layer for AI agents. Agents make autonomous decisions, hold tool permissions, and can take irreversible actions — none of which traditional scanners assess.

5 min read · Apr 10, 2026

Read