This talk shares practical lessons learned from hands-on testing of real-world generative AI application use cases, focusing on how security failures emerge when LLMs are integrated into production systems.

Attendees will learn:

• The most common risks identified in LLM-enabled applications and the guardrails that are frequently missing
• What “good” looks like for LLM guardrails, and how those guardrails can be evaluated in practice
• How to leverage guardrails in production as detection and response signals to defend against persistent attackers