Shared Security Podcast 
Tom Eston AI agents are not just chatbots. Once they can reset, approve, publish, delete, or change things, they need real security controls.
In episode 437, I discuss guardrails for AI agents: least privilege, read-only first, human approval, separate contexts, logging, and prompt-injection awareness.
Watch this episode on YouTube:
https://youtu.be/TL3MGnI4hUU
Listen and subscribe wherever you like to get your podcasts:
https://sharedsecurity.net/subscribe
https://sharedsecurity.net/2026/06/15/guarding-ai-agents-boundaries-and-safeguards/