Definition
An AI system that can independently decide what actions to take, invoke tools or APIs, and execute tasks without human approval between each step. Unlike chatbots that respond to user requests, autonomous agents plan, execute, and adapt their behavior based on results.
Why it matters
Autonomous agents represent a new class of system risk: they operate at machine speed, hold multiple credentials, and make decisions with real-world consequences. Traditional security models (requiring human approval for each action) break down at agentic scale.