Centaur Security Labs¶
Copyright © 2026 Jay Hawkins / Centaur Security Labs. All rights reserved.
Research and development in AI-augmented security operations — by Jay Hawkins.
Centaur Security Labs is a practitioner-led research lab built on a single question: what does principled human-machine collaboration actually look like in security operations — not in theory, but in production, against real targets, under real constraints?
The best security operations don't come from AI replacing analysts or analysts avoiding AI. They come from a principled three-layer division of labor: the model handles command generation, output interpretation, and technique selection — the adaptive execution layer. The code handles routing, safety enforcement, audit trail, and halt logic — deterministic behavior that can't be left to probabilistic inference. The human handles scope definition, organizational context, and accountability for decisions that carry legal or ethical weight. Each layer has a defined role. Violations of those boundaries — in either direction — are the primary reason AI security tools fail in production.
That model is the thesis of this lab. Everything here — the tools and the research — is built on it or derived from it.
The model in practice: ARCHER is a local-first AI penetration testing agent built on this architecture. The research documents what building and running it in production actually surfaces.
Articles and papers publish as they complete review. Subscribe via RSS to follow along.
The views expressed in publications on this site are those of the author and do not reflect the official policy or position of NORAD, USNORTHCOM, USCYBERCOM, the Department of the Army, the Department of War, or the United States Government.
Centaur Security Labs LLC — Colorado