Agentic security: Detection and response at machine speed

This article presents AWS's framework for securing autonomous AI agents at enterprise scale, developed in collaboration with SANS Institute.

• Agentic workloads operate autonomously with unpredictable behavior, requiring continuous detection and response at machine speed
• Only 10% of organizations govern AI despite 80% adoption, creating urgent security maturity gaps
• Agent identity and governance requires temporary, scoped credentials and zero trust principles for every request
• Continuous behavioral monitoring with adaptive baselines detects anomalies in real time, replacing static rule-based detection
• Automated tiered response distinguishes between actions that can be contained immediately and those requiring human judgment
• Multiagent ecosystems inherit all security requirements from single agents, requiring foundational security today
• AWS security services including GuardDuty, Inspector, and Security Hub adapt to agentic workload characteristics

AWS positions security as an enabler of responsible AI adoption by extending proven cloud security foundations to autonomous agent environments.