Sep 24, 2026 · 34 min
AI agents test the logic of cybersecurity controls
No factoring necessary.
As organizations embed AI into workflows, controls designed around human accountability must be reconsidered for autonomous systems.
- 1Segregation of duties was built to prevent fraud and assign accountability among humans.
- 2AI agents challenge whether traditional governance controls still work when systems act with limited human intervention.
- 3Organizations should match specialized AI tools to specific jobs instead of defaulting to the most powerful models.
Don't miss
Kurt Dusek explains why segregation of duties was designed for human fraud prevention and accountability, then tests whether those assumptions fit AI agents.
The brief
Enterprise AI is expanding faster than many organizations’ governance assumptions. Kurt Dusek of AppDome compares the moment with the early internet boom and examines the rise of agentic systems.
Segregation of duties was created to prevent fraud and make human accountability visible. AI agents complicate that logic because systems can perform tasks without the same incentives or responsibilities as people.
The discussion turns to AI adoption: organizations may be better served by specialized tools suited to defined jobs than by automatically choosing the most powerful general-purpose models.
The standout tension is whether controls built around human behavior can govern agents effectively, especially as products such as Claude and other AI systems enter enterprise workflows.