Broadcom unveils that top AI models from Google DeepMind, NVIDIA, NEC, Alibaba Cloud, and Z.ai can run on VMware Cloud Foundation, enabling on-premises AI deployment and private-cloud model-as-a-service.

AgentMinder provides an independent control plane for agent identity, enforcing least-privilege policies with a policy-driven gateway to approve or block actions and an OpenTelemetry-based audit trail.

VCF emphasizes data sovereignty and governance, allowing secure on-premises deployment and private-cloud AI workloads to keep data within enterprise infrastructure.

TrueSource portfolio adds clean-room builds and hardened images, supporting PostgreSQL, RabbitMQ, MySQL, and Valkey, with tiered site-license pricing disclosed but without figures.

Tanzu data foundations ingest, annotate, and govern structured, unstructured, and multimodal data to create governance-enabled data products published in the Tanzu marketplace, with RBAC and lineage.

Multitenant model sharing enables a single model to serve multiple business units while isolating tenant data.

Tanzu Platform acts as the agent layer with deny-by-default sandboxes, a secure development harness, a curated marketplace, and an AI gateway for monitoring and governance.

IDC Perspective underscores the need for unified governance, continuous authorization, and real-time telemetry so agent actions are observable, attributable, and reversible.

Cloud-native and deployable across Kubernetes environments, including VMware vSphere Kubernetes Service and Google Cloud Platform, with a multi-region active-active architecture to minimize downtime.

AgentMinder supports cloud-native deployment alongside on-premises, VPC, and public cloud LLMs, integrating with AuthZEN for policy enforcement endpoints and avoiding a single SaaS chokepoint.

AgentMinder integrates with existing enterprise authorization stacks via the AuthZEN standard, reusing current policy enforcement endpoints to avoid a single SaaS chokepoint.

Context notes a growing enterprise attack surface from agents, tools, datastores, and LLMs, underscoring the need to govern MCP and A2A communications with zero-trust principles.