en.Wedoany.com Reported - Trust3 AI has announced plans to extend its Unified Trust Layer to the NVIDIA NeMo platform, an end-to-end platform for developing custom generative AI models and agents. As a member of the NVIDIA Inception program, Trust3 AI is advancing this integration, aiming to enable enterprises to manage agents and applications running on NeMo without building additional independent control systems.
When organizations transition from pilot to production, the primary obstacle is no longer model capability but governance. Operations teams struggle to answer basic agent control questions: which agents are running, what data and tools they can access, what actions they actually perform, and whether this information can be audited and verified. Trust3 AI's planned NeMo integration aims to establish a governance closed loop at the action level.
Once integrated, it will provide agent discovery and inventory capabilities for teams building on NeMo, covering agents that may not be registered in NeMo deployments; implement purpose-bound access control, linking agent access permissions to their declared purpose rather than static roles, with automatic authorization expiration; enforce policies at the moment of operation execution, carrying agent identity, declared purpose, and data lineage information with each request before tool invocation or data movement; achieve full-fidelity observability, capturing prompts, retrievals, tool calls, and data access, supporting agent behavior replay and review; and generate audit evidence chains that can map to standards such as the EU AI Act and the NIST AI Risk Management Framework.
This approach complements NeMo's own control mechanisms, including NeMo Guardrails, by overlaying an independent, vendor-neutral trust layer. The trust layer sits beneath the agents, rather than being deployed in parallel.
Neeraj Sabharwal, co-founder of Trust3 AI, stated that when enterprises move from experimentation to production on NVIDIA NeMo, the main bottleneck is security and governance. This integration places the trust layer beneath NeMo-based agents, enabling security and compliance teams to observe every agent, protect and manage every operation, while ensuring AI delivery teams remain unaffected. Founded by core engineers of Apache Ranger and Apache Atlas, Trust3 AI's single control plane architecture and Unified Trust Layer are designed to manage agents across any framework, cloud, or data source.
