// BLOG
All posts.
Field notes from 20+ years in security. Plus the occasional detour into life, work, and whatever else is on my mind.
The Slow Call Is Still the Job
Security leadership rewards speed. The calls I am proudest of during an active incident were the ones I forced myself to slow down on. That is a trained discipline, not a personality trait, and most playbooks never write it down.
Your Vendor's Auditor Is Also a Vendor
SR Bancorp's breach came through the firm hired to audit its controls. Most vendor risk programs never ask who has access to the data behind the audit itself.
When Your Vendor Gets Breached, You Find Out Last
A fintech software vendor got hit with ransomware and up to 1.35 million banking customers were exposed. Most of their banks did the questionnaire, got the SOC 2, and still found out months late. The control that matters isn't the one you audited.
You Gave the AI Agent the Keys. That's the Breach.
Enterprise AI agent fleets doubled in four months and only about one in five are governed before they go live. When one gets breached, there's usually no clever attack behind it. Just an agent doing exactly what it was allowed to do.
Your AI Policy Doesn't Cover the Part That Can Actually Hurt You
79% of organizations are deploying AI agents. 6% have updated their governance to match. The gap isn't in the models. It's in what those agents are connected to.
The CISO Who Buries Bad News Isn't Wrong. The System Is.
95% of CISOs feel pressured to suppress compliance findings. The industry response is that they need more backbone. That's the wrong read.
Finding a Mentor Is Good Advice. Finding a Sponsor Is the Advice Nobody Gives You.
Mentors give you guidance. Sponsors spend their capital on you. Most career conversations focus on the first one. The second one is what actually moves careers.
Your Incident Response Plan Is Modeling the Wrong Threat Actor
LockBit dominated tabletops for years. The ransomware ecosystem has rotated. The groups hitting organizations right now are not the ones your IR team practiced against, and that gap has consequences.
Your AI Agent Has a Supply Chain. Did You Audit It?
One in four MCP servers expose AI agents to remote code execution. Most teams deploying agents do not know what an MCP server is. That is a supply chain problem disguised as an AI launch.
"We Have an AI Policy" Is the New "We Passed the Audit"
OpenAI just admitted prompt injection isn't getting solved, and companies are wiring AI agents into production anyway. A policy document is not a control.