The Road to Assume-Agentic, Assume-Inside

Two years of public cases: the threat got faster, AI-driven, and already inside; the federal response followed
FEB 2024
Nation-state · OT
Volt Typhoon
PRC state actor pre-positions inside US critical infrastructure using living-off-the-land tradecraft: staged to disrupt, not to steal.
2025
AI-orchestrated
GTG-1002
Anthropic reports the first documented large-scale AI-orchestrated cyber-espionage campaign: AI ran the great majority of the operation, humans only at key decision points.
DEC 2025 - FEB 2026
Single operator
Mexican multi-agency breach
One operator used Claude Code + GPT to breach nine agencies, with AI generating ~75% of commands: one person at the speed of a team. (as reported by Gambit Security)
APR 2026
Physical disruption
Iran: advisory AA26-097A
Iranian-affiliated actors manipulate PLC project files and HMI displays in US water, energy, and government facilities, real operational disruption.
APR - MAY 2026
The response
CISA responds
"Adapting Zero Trust Principles to OT" + CI Fortify: assume the adversary is inside; validate on identity; isolate and recover.
The trajectory is unmistakable: faster, AI-driven, and already inside. Defenses built for human-speed, perimeter-trust adversaries are aimed at the wrong threat; this playbook is built for this one.
Sources: Anthropic threat disclosure (GTG-1002); Gambit Security; CISA advisories & guidance (Volt Typhoon, AA26-097A, Adapting Zero Trust to OT, CI Fortify). Figures indicative: verify before quoting in formal settings.
© 2026 Ken Connell. All rights reserved.
AGENTIC SECOPS