Explore AI safety and security in this 52-minute DevSecCon episode featuring David Haber, co-founder of Lakera.ai and creator of Gandalf. Dive into prompt injections, AI behavior, risks, and vulnerabilities. Learn about data poisoning, protections, and the motivation behind Gandalf's creation. Examine two approaches to informing Large Language Models (LLMs) about sensitive data, discussing their pros and cons. Gain insights into LLM security and the importance of considering model-specific knowledge. Discover expert perspectives on AI safety, security, and play from David Haber in this informative discussion.
Overview
Syllabus
Ep. #137, AI Safety, Security, and Play
Taught by
DevSecCon