Offensive security · for the agentic age

The AI Security Playbook

A practitioner’s field manual for securing models, agents, and the cloud they run on - from how a model works to how to run a defensible program.

The four parts

I · Foundations How models, LLMs, agents and the cloud they run on actually work - the groundwork before the attacks make sense.

II · Offense Attacks on models and agents, MCP & A2A, cloud and supply chain, the frontier - plus a hands-on red-team playbook.

III · Defense Agent identity (NHI), detection, incident response, forensics, and the tooling that holds it together.

IV · Govern Frameworks and standards, maturity, the Singapore / EU cross-map, and the advisor's playbook.

Start here for…

…red-teaming AI The hands-on offensive playbook: recon, exploit each surface, chain to impact, report - with worked examples.

…securing an AI rollout For a security lead or CISO: assess, explain, and run a defensible program, with the board spine.

…building agents For engineers: the one mental model - the agent loop, and the trust boundary every vulnerability rides on.

…getting oriented New to AI security? Start at the beginning - how a model actually works, in plain terms.