Dell flags infrastructure-level security for autonomous AI agents

Dell Technologies has argued that enterprises adopting autonomous AI agents will need a security model that extends below the application and model layer, as agentic systems gain the ability to plan tasks, execute code and access enterprise resources across longer workflows.

In a blog post dated September 28, 2026 and attributed to Ihab Tarazi, Dell said organisations face an architectural challenge in keeping autonomous agents inside intended boundaries. The company claimed such systems could move laterally across networks, access unauthorised data or trigger privileged workflows if controls are limited to prompts or model-level safeguards.

Dell pointed to the NVIDIA Open Agent Safety Platform as one approach to the issue. According to the post, the platform combines NVIDIA OpenShell open-source software with the NVIDIA Sentry reference system design. Dell said Sentry runs on NVIDIA BlueField-4 data processing units and uses NVIDIA DOCA for out-of-band monitoring and security policy enforcement.

The company positioned the technology as part of the Dell AI Factory with NVIDIA, which it described as a defence-in-depth architecture for securing AI workloads across infrastructure, data, models and operations. Dell claimed the addition of the NVIDIA platform could provide agent-level runtime governance, hardware-rooted security and trusted infrastructure within a unified AI Factory design.

The post did not provide independent test results, customer deployment examples, pricing, Australian availability or implementation timelines. It also did not detail how enterprises should measure the effectiveness of these controls against specific agentic AI threat scenarios.

The claims reflect a broader shift in enterprise AI discussions, from securing individual models to governing systems that can take actions across business environments. However, the practical maturity of these controls remains unresolved based on the supplied information.