David Bombal

#539: Agentic AI is breaking your Cybersecurity controls (and how to solve it)

Feb 20, 2026
Guest
Peter Bailey
Peter Bailey, SVP and GM of Cisco Security with deep incident response and AI experience, explains how agentic AI is expanding the attack surface at machine speed. He outlines risks like shadow MCP servers, agents touching PII, and why perimeter controls fail. He also covers model provenance, agent identity and dynamic authorization, plus network controls like SD-WAN and SASE for AI traffic visibility.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Rapid Agent Adoption Then Pullback

  • Agent adoption exploded then halted as CISOs put 'shields up' after seeing risky behavior.
  • Unmanaged agents accessed PII and internal systems, expanding the attack surface at machine speed.
ANECDOTE

Developer Exposed LLM Accessing Sensitive Data

  • Peter described a recent incident where a developer unknowingly exposed an LLM and it accessed credit cards and sensitive data.
  • The story shows both agent power and the immediate security consequences of careless deployments.
INSIGHT

Agents Widen The Attack Surface

  • Agents widen attack surface by increasing entry points and exploiting seams between controls.
  • Purpose-built malicious agents will collapse attackers' cost curve and scale sophisticated attacks.
Get the Snipd Podcast app to discover more snips from this episode
Get the app