Broadcom moves security into hypervisor as agentic AI widens attack surface

The conversation at VMware Explore 2026 frames agentic AI not just as a workload change but as a structural challenge for enterprise security. Mahajan describes how customers have bought multiple security products that do not integrate, leaving seams that attackers exploit. Broadcom's answer is an integrated software stack sharing context across vDefend and Avi Load Balancer, pushing enforcement into the hypervisor where traffic inspection can occur without the delay of separate appliances. That design choice is grounded in the scale of AI workloads, which generate heavy east-west traffic and punish any added latency.

The emphasis on visibility reflects the transient nature of agents and Model Context Protocol services. Administrators need a real-time picture of what is authorized and what is shadow IT before they can quarantine anything, and that work is now being folded into private cloud modernization programs. Mahajan's warning is pointed: waiting two years to deploy lateral security means risking compromise, especially as automated attacks evolve. He frames the choice as deploying security at the infrastructure level and at scale now, rather than bolting it onto infrastructure already carrying production AI traffic.