The CISO Playbook for AI Agents | Datadog
Summary
A16Z’s Joel De La Garza interviews Datadog CISO Emilio Escobar on securing AI coding agents. They discuss managing credential access, using intent-based judges to detect malicious code, and avoiding restrictive security policies. Escobar argues that while AI accelerates vulnerability discovery, the real challenge is handling the resulting volume of alerts without panic.
Topics discussed
Intro: AI security risks and Datadog CISO Emilio Escobar
Datadog's strategy for widespread AI adoption
How AI agents bypass traditional data permissions
Securing credentials and sandboxing coding agents
Using AI judges to detect malicious code intent
The risk of AI agents solving problems destructively
CISO profiles and the shift in security talent
Bridging the gap between security teams and developers
Why Emilio is calm about AI 'going wild'
Concerns over vulnerability volume and AI hype
Listen ad-free on Castria