Acknowledgments

This book is the product of conversations, arguments, incidents, and lessons learned across dozens of organizations. I cannot name most of them—security work happens in confidence—but their fingerprints are on every chapter.

To the security architects who pushed back on my assumptions and forced me to think harder: thank you. The best ideas in this book emerged from those debates. The worst ideas were caught before they made it to the page.

To the incident responders who let me learn from their hardest days: thank you. Nothing clarifies architectural thinking like seeing what actually breaks.

To the ML engineers who patiently explained their systems without condescension: thank you. Security cannot protect what it does not understand. You made understanding possible.

To the teams who invited me to review their AI architectures and trusted me with what I found: thank you. Your willingness to be honest about gaps is rarer and more valuable than you know.

To the security leaders who gave their people time to do this work right, even when faster was easier: thank you. Architecture takes time. You created the space.

To my reviewers, who caught errors, challenged assumptions, and improved every chapter: thank you. The mistakes that remain are mine. The ones you caught would have embarrassed me.

To my family, who tolerated the late nights and distracted weekends that writing demands: thank you. Books are written in the time stolen from people who deserve it more.

And to the practitioners who will read this book and do the hard work of actually securing AI systems: thank you in advance. The ideas here are worth nothing until someone builds them. That someone is you.