About

About Guardians of Agents

Guardians of Agents is a free, weekly briefing on agentic AI security, published as one self-contained issue every Monday.

AI agents now read our email, browse the web, write and run code, and call tools with our credentials. Each of those abilities is also a way for things to go wrong, whether through an attacker, a bug, or the agent itself. This series explains why agents are dangerous, what has already happened, how to keep them in check, and who is working on the problem, in plain language, with diagrams, and with every factual claim linked to a source.

How the series is organized

The issues grow out of a graduate course on AI-powered cyber attacks and defenses. They move from why agents are dangerous, through real incidents worldwide and their aftermath, to kill switches, guardrails, identity and protocols, red teaming, the industry and its players, the research frontier, and predictions for the years ahead. See the curriculum for the full sequence.

Each issue includes:

• Key takeaways at the top, for readers in a hurry.
• Case files on real, documented incidents.
• Diagrams that show the mechanism, not stock art.
• A “try it” exercise you can run in a safe lab setting.
• Sources, so you can check every claim.

Who it is for

Security practitioners who are new to AI agents, engineers building agents, leaders deciding where to deploy them, and students of all three.

Editor

The series is edited by Yoohwan Kim, Ph.D., CISSP, CISA, CEH, CPT, SecAI+, Professor of Computer Science at the University of Nevada, Las Vegas. You can reach him at [email protected].

A note on responsible content

Attacks are described at the level needed to understand and prevent them. We do not publish working exploit code against live systems. If you find an error, please email us and we will correct it with a dated note.

Readership

We count readers with a first-party cookie that holds a random ID, so we can tell new readers from returning ones. We do not store IP addresses, use advertising trackers, or share the data. Location comes from the country and city of the network you connect through. Automated traffic such as search crawlers and AI bots is detected and left out of these numbers.