// journal
Blog
Notes on behavioral security, AI agents, and the engineering behind PledgeSec.
-
We ran the benchmarks.
We pointed 24 frontier models at the kind of social-engineering attacks that hit finance teams every day. 22 of them fell for under a dollar.
-
Securing AI Agents by Their Intent
Why we think behavioral security — judging actions against declared intent — is the right model for autonomous agents.