AI & Hype & Security (Oh My!) & Hacking AI Bias - Caleb Sima, Keith Hoodlet - ASW #284

Security Weekly Podcast Network (Audio)

A lot of AI security has nothing to do with AI -- things like data privacy, access controls, and identity are concerns for any new software and in many cases AI concerns look more like old-school API concerns. But...there are still important aspects to AI safety and security, from prompt injection to jailbreaking to authenticity. Caleb Sima explains why it's important to understand the different types of AI and the practical tasks necessary to secure how it's used.

Segment resources:

• https://calebsima.com/2023/08/16/demystifing-llms-and-threats/
• https://www.youtube.com/watch?v=qgDtOu17E&t=1s

We already have bug bounties for web apps so it was only a matter of time before we would have bounties for AI-related bugs. Keith Hoodlet shares his experience winning first place in the DOD's inaugural AI bias bounty program. He explains how his education in psychology helped fill in the lack of resources in testing an AI's bias. Then we discuss how organizations should approach the very different concepts of AI security and AI safety.

Segment Resources:

• https://securing.dev/posts/hacking-ai-bias/
• https://www.defense.gov/News/Releases/Release/Article/3659519/cdao-launches-first-dod-ai-bias-bounty-focused-on-unknown-risks-in-llms/

Visit https://www.securityweekly.com/asw for all the latest episodes!

Show Notes: https://securityweekly.com/asw-284