Stay informed

The rogue agent problem: A conversation with Gil Elbaz at the Hg Digital Summit

Orbit 63

In this episode, John Cranmer is joined by Gil Elbaz, Co-Founder and Chief AI Officer at Onyx Security, the company building the secure AI control plane for the agentic era. Gil, formerly in the CTO office at Nvidia working on multi-agent infrastructure, argues that the AI attack surface has fundamentally changed. The old boundary between what is yours and what is not has collapsed, and every agent, every MCP, every tool connection is now part of the perimeter. The conversation moves from the rise of shadow AI inside enterprises, where employees reach for any tool that makes work faster, to the harder problems underneath: rogue agents taking destructive actions they were never asked to take, multi-agent communication happening invisibly through Slack and email rather than direct A2A channels, and offensive AI systems like Mythos finding more zero-days in weeks than internal red teams find in years. Gil shares his playbook for getting your arms around it: visibility first, then policy, then runtime protection. As he puts it, the right seatbelt is what lets you drive faster.

Transcript

John Cranmer

Welcome to Orbit, the Hg podcast series where we talk to leaders and hear how they've built some of the most successful technology companies in the world. I'm John Cranmer, part of Hg's tech team, and my guest today is Gil Elbaz, Co-Founder and Chief AI Officer at Onyx Security.

Gil started in academia at Technion, was the CTO and co-founder at Datagen, creating data to train AI systems and pioneering synthetic data for visual AI. He then joined Nvidia, working as a direct report to the CTO on AI agents. Today, he's building Onyx, a secure AI control plane for the agentic era. Many organisations are now operating their own AI agents, and Onyx is helping them identify, monitor and control their usage.

Gil, great to have you.

Gil Elbaz

Thank you. It's a pleasure to be here. Thank you very much.