Jackson Reed built the red team at the US Army Corps of Engineers and took it through certification, then spent four years rebuilding and leading Target's. He isn't worried about swarms of agents overwhelming your SOC — defenders are good at bursty and anomalous. What concerns him is the operator with ten thousand actions a day of compute budget who spends ten of them, in each of a thousand environments, for a year.

We get into what a six-month timeline does to log retention and correlation windows, the two opposite ways red teams fail, why the frontier labs stopped measuring what matters in their system cards, and the Bronze Age analogy that explains most of the AI tooling being built right now.

TIMESTAMPS
0:00 Cold Open
1:14 Meet Jackson Reed
6:29 Are You Hopeful or Upset?
12:33 Patience Beats Zero-Days
21:21 Agents Criminals Don't Have to Trust
24:01 Red Teaming in a Hall of Mirrors
29:46 Can AI Clone a World-Class Red Team?
33:35 What's Actually Scalable with AI
38:05 One Frontier Model, or Ten Thousand Custom?
40:58 AI Overkill: When Less Is More
44:58 YOLO Security
46:08 Chinese Models and Who Has Your Data
52:05 Barding: Can You Handle an Intrusion?
53:32 What System Cards Stopped Measuring
1:02:35 AI's Bronze Age
1:07:34 Holding an Industry Hostage
1:10:44 Where to Find Jackson

ABOUT JACKSON
Jackson Reed studied international affairs before moving into computer science through the CyberCorps Scholarship for Service. He did think tank and SOC analyst work, then threat attribution and intelligence analysis at CrowdStrike, built and led the red team at the US Army Corps of Engineers through certification, and rebuilt and led Target's. He launched Barding Defense in March and works with BT6 on assessments and jailbreaking for frontier labs.

Blog: twelvetables.blog
Barding Defense: bardingdefense.com

#cybersecurity #redteam #aisecurity #threatintel #infosec #aiagents