One Operator, a Cast of Agents

Model the single operator's AI "workforce" as discrete agent-roles, then map each to a technique and a control
🤖
1 OPERATOR · an AI workforce
One person directing a cast of agents, running at the scale and speed of a full team, around the clock.
↓  DELEGATES TO  ↓
Recon / OSINTReconnaissance · TA0043
Automated target profiling, infrastructure mapping, and data-leak harvesting.
Defend: attack-surface monitoring; rate-limit and deny mass enumeration.
Social engineeringPhishing · T1566
Tailored lures, deepfake pretext, and multilingual phishing at scale.
Defend: inbound + deepfake anomaly detection; phishing-resistant auth.
Malware devResource Dev · TA0042
Generate, debug, and obfuscate payloads to evade signatures.
Defend: behavioral EDR + sandbox detonation; no signature-only reliance.
Infrastructure / C2Command & Control · TA0011
Spin up proxies, bulletproof hosting, and resilient C2.
Defend: egress / beaconing analytics; egress allow-listing; DNS controls.
Fraud / extortionImpact · TA0040
Negotiate, run money-mule logistics, and pressure victims.
Defend: BEC + anomalous-payment analytics; transaction holds.
ExfiltrationExfiltration · TA0010
Stage and move data faster than humans can react.
Defend: DLP + egress-volume baselines; egress caps and quarantine.
⚡
Flip the operator: the same cast, over-permissioned inside your org, is agentic insider risk. Govern each as a non-human identity. When the agents are yours, MITRE ATLAS techniques (prompt injection, tool abuse) turn them against you.
The defender's reframe: you don't need a control for "AI". You need a control for each function the workforce performs. Decompose, map to ATT&CK, and answer each role.
ATT&CK IDs are indicative, mapped at the tactic level except Phishing (T1566), a technique. Conceptual reference for an illustrative single-operator threat (APT_UNICORN).
© 2026 Ken Connell. All rights reserved.
AGENTIC SECOPS