AI agents are becoming a new identity attack surface across AWS, Azure, SaaS, and hybrid environments. Unlike traditional software, AI agents operate through identity chaining—dynamically inheriting IAM roles, service accounts, delegated tokens, and connector permissions at runtime.
In this webinar, experts from Silverfort explore why traditional IAM, PAM, and deterministic security models break down in agentic AI environments—and how attackers are already exploiting AI agents through prompt injection, delegated token abuse, privilege escalation, and agent-driven lateral movement.
Topics include:
- AI agents as delegated and composite identities
- Identity chaining across AWS Bedrock, Azure OpenAI, and Copilot Studio
- Prompt injection, context poisoning, and instruction override
- Ghost permissions, connection leakage, and privilege sprawl
- AI agents as a new lateral movement bridge
- Why modern attacks target intent—not just credentials
- Runtime identity protection and inline enforcement for AI agents
- MITRE ATT&CK alignment for AI agent threats
This session breaks down how AI agents inherit permissions, traverse cloud environments, and create new identity risks that traditional controls were never designed to stop.
Learn how to secure AI agents with identity-first, inline protection that operates at machine speed—before authorized behavior becomes unauthorized impact.