Red Team vs Blue Team - Securing AI Agents

Explore the security landscape of AI agents in this conference talk that examines the evolution from traditional APIs to autonomous agents and their associated security challenges. Discover real-world adversarial threats including prompt injection attacks, tool abuse scenarios, and agent-to-agent escalation vulnerabilities that pose risks to production deployments. Learn comprehensive defensive strategies through practical implementation of rate limiting mechanisms, encrypted endpoint configurations, and Amazon Bedrock Guardrails to protect against malicious activities. Master the principles of building secure-by-default AI agents using Infrastructure as Code methodologies that enforce deterministic behavior patterns and implement Zero Trust security frameworks. Gain essential knowledge for securing autonomous agent deployments in production environments, with hands-on demonstrations of security controls and best practices for teams responsible for AI agent infrastructure and operations.