Tanium extends endpoint management to NVIDIA AI kit

Tanium has teamed up with NVIDIA to extend endpoint management to NVIDIA local AI infrastructure, including RTX Spark, DGX Spark, DGX Station and DGX Station for Windows.

The move brings Tanium's management tools to a class of AI devices that has often sat outside standard corporate IT processes because of its mix of hardware and operating systems.

Enterprises adopting NVIDIA's local AI systems will be able to use the same management approach across those machines as they do across other endpoints in their estate, with visibility into inventory, running processes, patch status and compliance.

A key part of the announcement centres on NVIDIA DGX Spark, which runs on NVIDIA DGX OS, an Ubuntu-based Linux distribution designed for AI workloads. That puts it outside the Windows and x86 environments on which many enterprise management products have traditionally focused.

The Tanium Autonomous IT Platform is listed in NVIDIA's DGX Spark Enterprise Lifecycle Integration Guide, and has been validated on NVIDIA DGX OS. According to Tanium, patching, asset inventory, configuration compliance and process management work without a custom build.

Support already covers DGX Spark and RTX Spark and is being extended to DGX Station and DGX Station for Windows.

Management gap

The companies are targeting a shift in AI computing from centralised data centre infrastructure to deskside and local systems used by developers and employees. That change has created new management challenges for IT teams, particularly when local AI devices fall outside established endpoint oversight.

Tanium says it can apply real-time query functions, natural-language workflows and policy controls across fleets of AI systems. It also says the platform is among the first endpoint management products to govern on-device AI agents at deployment and during runtime.

Highlighted functions include fleet-level consumption monitoring, giving operators visibility into which models are running and how much compute and memory they are using. IT and security teams can also start, monitor and stop AI agents automatically when they cross defined time or resource thresholds.

Tanium says its deployment tools can distribute AI agents as packages or processes across an entire fleet at the same time, and that its architecture is designed to scale from dozens of units to thousands.

Other features include visibility into CPU, memory, GPU and temperature data across fleets. Tanium also pointed to CIS benchmark hardening, compliance reporting mapped to CIS and NIST frameworks, software bill of materials tracking across packages, containers and language runtimes, and privileged access controls with session recording.

Air-gapped deployment support is also available for regulated environments.

Executive views

Harman Kaur, Chief Technology Officer at Tanium, said the spread of AI computing into enterprise endpoints had introduced new operational demands for IT and security teams.

"NVIDIA's vision for AI compute is bringing a new class of endpoint into the enterprise, including new hardware architectures, new operating systems, and AI agents running locally on all of them," Kaur said.

"To manage that infrastructure with confidence, IT and security operators need real-time visibility and control across their entire fleet. Tanium extends that capability to NVIDIA's AI endpoints the same way we have extended it everywhere else. With attacks now moving at machine speed, there's no room for blind spots," Kaur said.

NVIDIA framed the tie-up as a way to connect local AI hardware with the management processes companies already use elsewhere in their organisations.

"NVIDIA DGX and RTX platforms bring powerful local AI to enterprise developers and employees, enabling them to build and run AI within their established workflows," said Chris Marriott, Vice President of Enterprise Platforms at NVIDIA.

"Our collaboration with Tanium extends that integration to IT operations, helping organizations incorporate these platforms into fleet management processes they already use as they deploy AI compute at scale," Marriott said.

The collaboration points to a broader effort by infrastructure and management software suppliers to adapt endpoint oversight to AI systems that are moving closer to employees rather than remaining within dedicated data centre environments.