
HiddenLayer: Chris Sestito on ML/AI Security Incidents and Defense Capabilities
Secure Ventures with Kyle McNulty
00:00
Unveiling AI Vulnerabilities Through Inference Attacks
Exploring a significant AI security incident revealing vulnerabilities in a multi-billion dollar company's AI technology through an inference attack. The chapter discusses how attackers manipulated the model's least confident decisions to create a universal bypass, shedding light on the real-world consequences and motivations behind such breaches.
Transcript
Play full episode