As AI agents become more autonomous, generating software on-the-fly from human prompts, one question looms larger than ever: how do we keep them secure? In this episode of Invisible Machines, Robb Wilson and Josh Tyson sit down with Omar Santos, Distinguished Engineer of AI Security at Cisco and co-chair of the Coalition for Secure AI, to explore the evolving landscape of AI security in the agentic era.
Omar argues that traditional security models are no longer sufficient. The idea of a security department feels both antiquated and woefully inadequate. As AI agents create complex software environments dynamically, security must become an ever-present, integrated layer, supported by constant human oversight and the ability to simulate potential outcomes to mitigate risk. For organizations racing toward AI adoption, ignoring security isn’t just risky, it’s a barrier to progress.
The conversation dives deep into how AI agents are transforming work, teams, and technology ecosystems. Omar explains how advanced orchestration combines human judgment with AI capabilities, and why simulations and real-time risk assessments will be critical as agents evolve. He also shares insights from his work leading AI security at Cisco and guiding industry standards like CSAF and VEX.
For anyone exploring agentic AI, this episode is a masterclass in responsible innovation. It challenges leaders to rethink security as a core part of AI design, adoption, and management, because in the age of agentic AI, security is fundamental.
