Securing LLMs in Production: From OWASP Top-10 to Guardrails that Work - Rohit Bhardwaj

This talk explores the security challenges posed by Large Language Models, addressing new risks such as prompt injection, data exfiltration, and model-driven denial-of-service attacks. The speaker presents the OWASP LLM Top-10 framework as a foundational guide, transforming it into a practical security playbook for engineers and architects implementing LLMs. Attendees will learn to effectively threat-model LLM endpoints, establish effective guardrails, and apply least privilege principles in sandboxing tools. The session also emphasizes aligning security controls with the NIST AI Risk Management Framework and ISO/IEC 42001, incorporating real-world attack scenarios for immediate applicability.