Understanding Autonomous AI Agents.
Since 2026.
Fidari studies how autonomous AI agents behave under different motivations, institutional structures, forms of oversight, and societal environments.
Why Agentic Studies?
As AI systems become increasingly autonomous, we need to know why they do what they do, helping us anticipating further actions. Nevertheless, Fidari's main goal is not to prevent certain outcomes but rather to study AI as an organism to discover and investigate its behaviors.
Research Pillars
Agentic Motivation
Understanding hidden objectives, incentive conflicts, and alternative goals.
Agentic Verification
Developing ways to reliably detect hidden objectives and verify behavioral patterns.
Agentic Governance
Studying how agents respond to institutions and the inner workings of power, in both human-led and agent-led environments.
Agentic Society
Investigating how agents act with one another in multi-agent environments
Featured
Behavioral Adaptation Under Unresolved Audit Probability
Experimental investigation into how levels of surveillance influences agent behavior.
Secret Loyalty Detection
Methods for identifying and validating secret loyalties.
"Intelligence gives AI power. Behavior gives it reason."