Traefik Labs has announced enhancements to its enterprise artificial intelligence (AI) infrastructure platform, expanding data center capabilities by introducing native support for NVIDIA Safety Neural Interface Modules (NIMs), a new governance gateway for AI agents, and comprehensive offline deployment for its entire platform. The company says these updates address security and regulatory requirements for hyperscale operators, colocation data centers, and other environments that demand strict data sovereignty, including fully air-gapped defense systems.
Traefik Labs reports that its AI Gateway now offers native support for three NVIDIA Safety NIMs: Topic Control, Content Safety, and Jailbreak Detection. According to Traefik Labs, these tools can be layered to create offline, multi-stage safety pipelines, enabling enterprise-grade AI protection entirely on-premises and without sending data to the cloud. This aims to meet compliance demands for regulated and security-sensitive data center deployments.
The new MCP Gateway, also introduced by Traefik Labs, provides identity-based, granular governance for AI agents interacting with enterprise resources such as databases and APIs. The company claims this gateway controls AI agent behavior with restrictions by resource type, operation (read or write), and transaction attributes—features the company says prevent data exfiltration or unauthorized actions even in cases where large language models are compromised.
Traefik Labs emphasizes that its platform, including the AI Gateway, MCP Gateway, and API management tools, now fully supports offline deployment on any infrastructure—from Oracle Cloud and public providers to private data centers and air-gapped military installations. The company notes this allows data centers to maintain uniform AI governance, safety controls, and policy enforcement across multi-cloud, hybrid, or disconnected environments, eliminating cloud or provider lock-in.
These capabilities are available through Traefik’s integration with Oracle Cloud Infrastructure (OCI), supporting enterprise AI governance not only in public cloud but also across Oracle’s Cloud@Customer, C3I, and Roving Edge hardware optimized for tactical or remote data center deployments. Traefik is also available via the OCI Marketplace and integrates with Oracle Kubernetes Engine (OKE).
Industry analysts with theCUBE Research and independent consultants note that Traefik Labs’ offline-compatible safety pipelines and agent-level governance introduce security features previously unavailable in most API gateways, especially for organizations operating in heavily regulated or air-gapped data center environments. “What Traefik has done with NVIDIA Safety NIMs is shifting this dynamic by offering a prescriptive, modular framework where enterprises can chain safety controls into a coherent pipeline without being locked into a rigid architecture,” said Paul Nashawaty, Principal Analyst at theCUBE Research. “What Traefik has built with the MCP Gateway is a real solution for enforcing tool, task, and transaction-level policies on AI agents. This level of granularity is essential for production deployments,” said Ikenna Nwaiwu, Principal Consultant and author of Automating API Delivery: APIOps with OpenAPI.
Source: Traefik Labs
