This is what version 2 looks like
Version 2 of an agent-orchestrated OT/ICS fuzzing lab: two hosts, two architectures, an hourly rotation, honest validation tiers, and a human at the gate. The orchestration is the point.
Read the essayAI security · Vulnerability research · Digital forensics
Drew Stelly, PhD
I study how systems fail, how to test them, and how to make the results useful.
Security researcher working across enterprise AI, agentic offensive AI orchestration, and reproducible forensic analysis.
01Upcoming book
In developmentOrchestrating
the Attack
Agentic AI and the New Operating Model for Offensive Security
What happens to offensive security when the operator’s job becomes orchestrating teams of AI agents?
Orchestrating the Attack examines the changing role of human expertise, the economics of agentic security, and the challenge of producing evidence people can trust. Drawing on experience in red teaming, AI security, and digital forensics, Drew Stelly explores what this shift means for operators, defenders, security leaders, and AI researchers—and the judgment and accountability it demands.
Book inquiries02Selected writing
Essays & research notesThe numbers aren't the point. The orchestration is.
Pointing the same agent orchestration model at OT/ICS protocol stacks. Early receipts from the lab campaign — and why the loop matters more than the crash counter.
Read the essayBeyond security.txt: When the Reporter and the Triager Are Both AI
What vulnerability disclosure needs when software can both submit and review reports. A proposal for machine-readable intake, clearer evidence, and better triage.
Read the essayOrchestrating AI Agents for Vulnerability Research
Lessons from coordinating AI agents for vulnerability research: shared context, evidence quality, the limits of automation, and where human judgment matters.
Read the essay- 2025
Putting the ‘I’ in CIA for AI Models: A Framework for Model Integrity
Model integrity · January 6, 2025 · Cyber AI Guy
- 2025
LLM safety and C. S. Lewis
AI safety · February 10, 2025 · Cyber AI Guy
More notes on adversarial machine learning and AI security.
Read the blog03Publications
2015–2019Languages and tools for making security analysis reproducible. Academic work published as Christopher D. Stelly.
- 2019Read
A Domain Specific Language for Digital Forensics and Incident Response Analysis
PhD dissertation · University of New Orleans
A language and execution environment for specifying, recording, and reproducing forensic computations.
- 2019DOI
Language-based Integration of Digital Forensics & Incident Response
Christopher Stelly & Vassil Roussev · ARES 2019 · ACM
Connecting forensic tools and incident response through a shared language.
- 2018Read
Nugget: A Digital Forensics Language
Christopher Stelly & Vassil Roussev · Digital Investigation
Expressing digital forensic investigations as executable, auditable computations.
- 2017Read
SCARF: A Container-Based Approach to Cloud-Scale Digital Forensic Processing
Christopher Stelly & Vassil Roussev · Digital Investigation
Using containers to scale digital forensic workloads across cloud infrastructure.
- 2016
Red Crawling for Perimeter Intelligence
Lockheed Martin Journal of Software Engineering · Internal publication
- 2015
Dynamic Bytecode Instrumentation for Mobile Security
Lockheed Martin Journal of Software Engineering · Internal publication
04Background
Research in practiceMy work connects hands-on security testing with research that other people can use. A current focus is agentic offensive AI orchestration: how specialized agents coordinate, share context, and produce reliable evidence for security assessments.
I built the AI Red Team function at Thomson Reuters, following nearly a decade in security engineering and red teaming at Lockheed Martin.
2022–Present
Thomson Reuters
AI Security Researcher
AI security assessments, research, and methodology across language models, agentic systems, and the applications around them.
2012–2022
Lockheed Martin
Associate Fellow · Staff Red Team Operator
Security engineering and red teaming across enterprise and defense systems, with research in embedded systems and RF security. NOVA Award recipient.
05Selected talks
- 2023
Attacking AI — Stories and Secrets
Thomson Reuters · Fall 2023
- 2023
AI and Trust
Thomson Reuters · Spring 2023
- 2021
Red Teaming with Radios — Cyber RF/EW Testing
Lockheed Martin Fellows Conference
- 2020
A Distributed C2 Framework — DVN
Wild West Hackin’ Fest