AI Infrastructure Firms Strengthen Security as AI Factories Scale

SiliconANGLE published interviews and feature reporting that place AI factories, integrated, end-to-end stacks for production AI, at the center of a widening security conversation. Per SiliconANGLE, a Dell Technologies executive identified by surname as Khatri, described LLMs and model weights as sensitive assets that "need to be protected like code" and warned that AI brings "additional threats, newer threats" during an interview on theCUBE, SiliconANGLE reports. The same coverage flags attacker techniques such as "living off the land," where trusted tools and agent workflows are repurposed by adversaries, and highlights lineage, provenance and observability as emerging baseline requirements for secure AI deployments.

A Dell blog post authored by Scott Bils, Vice President of Professional Services at Dell, presents an architectural approach to "Securing the AI Factory," arguing for security controls embedded across data, compute, and operational layers, per the Dell blog dated October 30, 2025. Separately, PR Newswire reports that Trend Micro announced a co-developed OEM appliance offering with Dell and NVIDIA aimed at delivering pre-validated, pre-integrated secure infrastructure for AI at scale; the PR quotes Kevin Simzer, COO at Trend Micro, on automating and simplifying security operations across cloud, hybrid, and air-gapped environments.

Industry reporting frames the security challenge as structural rather than incremental. Observed patterns in comparable enterprise-scale transitions show that when workloads change from batch analytics to continuous, stateful AI pipelines, tooling that was once optional becomes mandatory. For practitioners, this typically means deploying stronger provenance and lineage systems, runtime observability for model and agent behavior, and stricter identity and access controls for nonhuman principals. These requirements tend to increase operational complexity and drive demand for integrated or co-engineered solutions that bundle hardware, software, and managed services.

Public reporting highlights three drivers raising the stakes for security teams: models as secret artifacts, distributed and transient data flows across cloud/on-prem/edge, and the rise of autonomous agents that can act on behalf of systems. This combination expands the attack surface and changes the threat model in ways that legacy perimeter defenses were not designed to handle. Vendor collaborations and OEM appliances, like the Trend Micro/Dell/NVIDIA offering reported by PR Newswire, aim to lower integration friction for enterprises that lack deep in-house security and AI operations expertise.

• •Adoption of integrated provenance and lineage tooling across major cloud and on-prem platforms
• •Productization of least-privilege identity controls for agent and model execution
• •New compliance and audit controls tailored to model artifacts and data transformations

Industry observers and security teams will be able to track vendor roadmaps, co-engineering announcements, and standards activity to see which approaches gain traction.

This piece synthesizes SiliconANGLE reporting and interviews (theCUBE broadcast excerpts), a Dell blog post by Scott Bils, and a PR Newswire announcement from Trend Micro. Where quotations or firm announcements appear, they are attributed to the original publisher. Editorial commentary here is labeled and framed as industry-level observation rather than claims about any company's internal intentions.