Defending against indirect prompt injection in agentic ai

Created using ChatSlide

Explore the concept of indirect prompt injection, its risks, and relevance to AI systems. Understand emerging threats, real-world vulnerabilities, and defence mechanisms such as provenance tracking and runtime monitoring. Gain insights from research on detection and system robustness, along with innovative security measures like context-aware guardrails and secure-by-design principles.