AI Security Fundamentals: The Real Industry Shift Taking Place
Varonis, Friday, September 25th, 2026
Varonis says 'security for AI' isn't a separate discipline, and AI adoption often exposes pre-existing security gaps.
In a conversation between cybersecurity leaders Mike Privette and former Varonis Field CTO Matt Lock, Varonis argues that securing AI systems isn't a distinct discipline but requires the same governance, access controls and monitoring as the rest of the technology stack.
Adopting AI often functions like an unplanned penetration test, exposing existing weaknesses such as excessive permissions, weak identity controls and poor data classification.
Early AI security efforts narrowly focused on protecting chatbots and models, but the industry has shifted toward using AI to strengthen cybersecurity broadly. The piece compares this trajectory to cloud computing, predicting AI use will eventually be assumed rather than marketed as distinctive.