The rapid adoption of AI agents and LLM-powered applications is introducing a new generation of security challenges. Traditional network security controls were designed for applications, users, and workloads, but AI-driven systems introduce new attack surfaces involving prompt injection, tool abuse, data exfiltration, insecure MCP integrations, and unauthorized agent actions.

In this session, we will explore how to rethink network security for the AI era. We will examine emerging attack techniques against LLM applications and autonomous AI agents, understand why traditional security controls are insufficient, and explore how AI firewalls and modern network security architectures can protect AI workloads.

Through practical demonstrations and real-world attack scenarios, attendees will learn how to identify vulnerabilities, implement security controls, and build defense-in-depth architectures for AI\-powered applications.

Key topics covered:

- Emerging security threats targeting LLM applications and AI agents
- Direct and indirect prompt injection attacks
- MCP security, tool abuse, and excessive agent permissions
- Data exfiltration and unauthorized AI actions
- AI firewalls and runtime security controls
- Integrating network security with AI application security
- Building defense-in-depth architectures for AI workloads