Broadcom pitches an “end-to-end” security package for agentic AI at VMware Explore

Broadcom said it is rolling out a set of capabilities aimed at securing “agentic AI,” a term used for AI systems that can take actions toward goals through autonomous or semi-autonomous behavior. The company’s pitch, presented around VMware Explore 2026, centers on what it calls end-to-end coverage spanning security, identity, and observability, rather than treating those functions as separate tools.

In the announcement, Broadcom described AgentMinder as a component intended to help manage and control autonomous agents. While the release does not provide detailed technical specifications in the information available here, it frames AgentMinder as part of a broader security stack designed for organizations deploying agentic AI in production settings.

Broadcom also pointed to VMware vDefend, a product name used by VMware for security-related capabilities. In the company’s framing, vDefend is positioned as a layer in the stack that addresses threats and risk as agentic systems interact with enterprise environments. The release information available does not specify which vulnerabilities or classes of attacks the offering is tuned for, nor does it describe performance or deployment requirements.

For network and application traffic, Broadcom highlighted the VMware Avi Load Balancer. Load balancers generally sit in front of services to distribute traffic and can be used to enforce policies at the network edge. In this case, Broadcom tied Avi to the goal of supporting security coverage for agentic AI workloads, but it did not provide granular details on policy controls or how the load-balancing layer is integrated with agent activity.

The combined message is that enterprises will need security controls that align with how agentic AI systems operate, including the fact that agents and their supporting services can create dynamic interactions across systems. Broadcom’s wording in the announcement emphasizes securing not just agents themselves, but also “protocols” and “AI workloads,” suggesting a focus on both the communications that agents use and the environments in which AI workloads run.

Broadcom positioned the effort for enterprises building or expanding agentic AI deployments, a market where CIOs and CISOs face heightened scrutiny about identity and access, auditability, and visibility into system behavior. Security and observability are especially important in agentic environments because it can be harder to predict what actions a model-assisted system will take once it is connected to internal tools.

Still, the public details in the announcement information available here are limited. It does not disclose pricing, packaging, general availability timing, or whether the components require new licenses or upgrades from existing VMware customers. It also does not outline any measurable outcomes, such as reduced time-to-detect incidents, improved policy enforcement, or specific compliance mappings.

What to watch next is whether Broadcom will publish deeper technical documentation that clarifies how AgentMinder, vDefend, and Avi work together, including any integration points with identity systems and how observability is implemented for agent actions. Enterprises considering agentic AI deployments will also want clarity on implementation steps and any operational impacts, which were not provided in the available announcement excerpt.