TrendAI has built Nvidia’s open Nemotron models into Vision One for Sovereign and Private Cloud (SPC), and the key change is where the intelligence runs. AI features such as guided investigation and agentic SIEM now operate entirely inside cloud environments the customer manages. The SPC version is available immediately.
Until now, organizations that refused to send security data to a public cloud service often faced an uncomfortable trade. They could have AI-driven security operations or complete control, but rarely both. TrendAI says this release removes that choice. Vision One SPC runs in private clouds, sovereign clouds and on-premises environments, so the model travels to the data instead of the reverse.
Customers decide where the model runs, which version they deploy and who gets access. No external cloud inference service is required. The models run on Nvidia hardware in the customer’s own data centers, in sovereign clouds, or in environments with limited connectivity. That last detail matters for organizations that keep sensitive networks deliberately isolated from the public internet.
In practice, Nemotron analyzes security signals to spot complex threat patterns, attack chains and AI-native attack paths across cloud workloads. Organizations set the policies that trigger protective actions. The system also helps analysts investigate threats and decide what to tackle first. Meanwhile, the agent-based SIEM correlates telemetry while the security team keeps its overview. A partnership with Nvidia on OpenShell adds more control over, and visibility into, what the agents actually do.
TrendAI CEO Rachel Jin framed the release around a wider definition of sovereignty. Where data sits is only part of it, she said. Enterprises also need control over where the intelligence acting on that data operates and how it is governed. That argument fits cloud buying habits well, since many regulated buyers now ask where a provider’s AI runs, not just where its storage sits.
Buyers should still ask practical questions. The announcement does not list hardware requirements, pricing or performance figures, so teams will want those before planning a rollout. Running models locally also means the customer owns capacity, patching and monitoring for that private cloud stack.
For regulated industries and government bodies, though, keeping both the data and the AI inside the fence could ease a long-standing argument with security and compliance teams.
