Trust3 AI Plans to Extend Its Agent Control Plane to NVIDIA NeMo

Building on its NVIDIA Inception membership, Trust3 AI intends to bring agent discovery, purpose-bound access control, and full-fidelity observability to generative AI applications built with NVIDIA NeMo.

Trust3 AI, the agent control plane for the enterprise, announced its intent to extend its Unified Trust Layer to NVIDIA NeMo, an end-to-end platform for developing custom generative AI models and agents. As a member of the NVIDIA Inception program, Trust3 AI is working toward an integration designed to let enterprises govern agents and applications running on NeMo without standing up a separate set of controls.

As organizations move from pilots to production, the gap they hit is rarely model capability.It is governance. Teams struggle to answer basic operational questions: which agents are running, what data and tools they can reach, what they actually did, and whether any of it can be proven to an auditor. Trust3 AI’s planned NeMo integration is intended to close that gap at the point of action.

Once available, the integration is intended to give teams building on NeMo:

• Agent discovery and inventory across NeMo-based deployments, including agents that would otherwise go unregistered (Shadow AI).
• Purpose-bound access control (PBAC), where an agent’s access is tied to its declared purpose rather than a static role, with auto-expiring grants.
• Policy enforcement at the moment of action – before a tool fires and before data moves carrying agent identity, declared purpose, and data lineage with every request.
• Full-fidelity observability, capturing prompts, retrievals, tool calls, and data access so an agent’s behavior can be replayed and reviewed.
• Audit evidence mapped to frameworks such as the EU AI Act and the NIST AI Risk Management Framework.

Also Read: AiThority Interview with Matej Bukovinski, Chief Technology Officer at Nutrient

The approach is intended to complement NeMo’s own controls, including NeMo Guardrails, by adding an independent, vendor-neutral trust layer that sits beneath the agents rather than alongside them.

“As enterprises move from experimentation to production on NVIDIA NeMo, the gap they hit isn’t capability,  it’s security & governance,” said Neeraj Sabharwal, Co-Founder of Trust3 AI. “Our intent is to put the trust layer directly under NeMo-based agents, so security and compliance teams can see every agent, secure & govern every action, and prove it  without slowing the teams shipping the AI.” Neeraj Sabharwal, Co-Founder, Trust3 AI

Trust3 AI was founded by the engineers behind Apache Ranger and Apache Atlas, the open-source access-control and metadata standards used by regulated enterprises today. The company’s One Control Plane architecture and Unified Trust Layer are built to govern any agent, on any framework, across any cloud and data source.