Enterprise infrastructure software is being redrawn around a user that never sleeps, never logs off and can act thousands of times a minute. As autonomous agents move from pilots into production, the controls built for human employees are proving a poor fit for machine identities operating at speed.
That shift is turning the network and identity layers into the front line for agentic deployments, a theme running through much of this year’s VMware Explore coverage. Securing those workloads has to start before the first agent ships, according to Umesh Mahajan (pictured, left), vice president and general manager of the Application Networking and Security division at Broadcom Inc.
“If you don’t have security, you really can’t go deploy agentic AI,” said Mahajan. “First, you have to make sure that the agentic AI workloads don’t get compromised from the outside. Then, we see the agentic AI workloads can themselves go rogue and start attacking outwards, so you need security both inwards and outwards.”
Mahajan and Clayton Donley (right), vice president and general manager of the Identity Management Security division at Broadcom, spoke with theCUBE’s Christophe Bertrand and co-host Alison Kosik at VMware Explore, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. The discussion centered on how enterprises can secure agentic AI by combining network-level visibility, machine identity, policy enforcement and observability across Broadcom’s VMware Cloud Foundation security stack. (* Disclosure below.)
Identity and network controls converge in infrastructure software
Decades of access management were designed for people, and that inheritance does not transfer cleanly to software that acts on its own behalf. Broadcom’s answer includes AgentMinder, which pairs signed agent identities with runtime inspection of application programming interface traffic and observability, Donley explained.
“We’ve had 50 years of figuring out how to manage your employees or customers or other people accessing your computer systems,” Donley said. “We’ve had about 15 minutes to figure out how to do it for AI agents. The problem is that we’re giving these agents more and more autonomy, more ability to do things on their own without those controls.”
Sitting in the packet path lets the network tier discover Model Context Protocol servers, agents and models, then flag unauthorized ones as shadow AI, Mahajan noted. Putting that together from separate products is where customers stumble, which is why he advocates for a single infrastructure software stack.
“If we make a customer stitch together multiple products, they have an awful time; they make mistakes,” Mahajan said. “Having a curated security stack completely from Broadcom goes a long way in securing them.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of VMware Explore:
(* Disclosure: TheCUBE is a paid media partner for the VMware Explore 2026 event. Neither Broadcom, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)
