What Makes a Good AI Security Deployment?
Summary
Hosts David Spark and Eddie Contreras welcome guest Peter Liebert to discuss the lack of mature AI security reference architectures. They argue that AI security is an operational discipline, not a product, requiring cross-functional governance and clear ownership by the CISO. The episode highlights the need to distinguish between AI safety and security, simplify complex controls into data, access, and trust, and avoid paralysis by starting with small, measurable improvements.
Topics discussed
Intro: The lack of mature AI security reference architectures
Vendor transparency issues and undefined AI terms
AI as operational discipline vs. product; Simplifying controls
Paralysis by analysis and the 'eating the elephant' problem
Business literacy gaps and vendor hype in AI security
Vendor confusion and the need for unified workspace security
Cross-functional governance and CISO ownership of AI security
Improving communications and finding quick wins via employee feedback
Workflow optimization before applying AI solutions
Distinguishing AI safety from AI security controls
Favorite quotes, hiring at SalesLoft, and show outro
Listen ad-free on Castria