TLDR: Xage Security has announced the integration of its Xage Fabric Platform with Nvidia’s BlueField data processing unit (DPU) to extend zero-trust security to AI agents and data centers. This collaboration aims to provide advanced security controls for ‘AI factories’ and critical infrastructure, ensuring identity-based access controls, preventing data leakage, and safeguarding AI workloads at unprecedented scale and speed.
Xage Security Inc., a leading zero-trust cybersecurity company, revealed on October 28, 2025, a significant advancement in its security offerings through the integration of its Xage Fabric Platform with the Nvidia BlueField data processing unit (DPU). This strategic integration is designed to deliver advanced security controls specifically tailored for emerging ‘AI factories’ and critical infrastructure environments.
The core objective of this collaboration is to scale identity-based access controls across vast data centers and mission-critical settings, promising exceptional performance and resilience. The integration is particularly crucial for agentic artificial intelligence systems, which are characterized by autonomous agents collaborating with other agents, models, and application programming interfaces (APIs) to make decisions and execute actions.
Xage’s dynamic access control, powered by Nvidia BlueField, establishes a software-defined, hardware-accelerated foundation. This foundation is engineered to govern and enforce data flows between AI agents and models in real-time. The system implements ‘least-privilege’ controls at every stage of these interactions, meticulously dictating which agents can access specific data, pipelines, or models, as well as the precise actions they are permitted to perform and for what duration.
Organizations leveraging this integration can effectively prevent unauthorized privilege escalation and data leakage through role-based segmentation, which operates at line speed directly on the BlueField DPU. Furthermore, the platform enables the enforcement of policy-based privilege de-escalation, blocking risky actions and thereby ensuring that AI agents maintain trustworthiness and compliance as they evolve and scale. This capability is vital for enterprises seeking to harness the full productivity of agentic AI without incurring undue risk, even in the most complex and autonomous operational landscapes.
Ofir Arkin, senior distinguished architect for cybersecurity at Nvidia, emphasized the importance of this development, stating, “As AI factories emerge as the foundational infrastructure accelerating AI innovation, safeguarding them has become a critical priority. Together, Nvidia BlueField and Xage’s zero-trust security enable organizations to modernize their protection strategies across AI factories and infrastructure — driving secure, scalable innovation forward.”
Beyond access control, Xage also enforces the secure separation of AI workloads, a critical function that prevents lateral movement of threats and comprehensively safeguards datasets, workloads, and models. The direct implementation on the BlueField DPU ensures that Xage’s identity-based segmentation operates with low-latency and high-efficiency, effectively reducing the attack surface without compromising performance.
This integration is particularly relevant for data centers and critical infrastructure across various sectors, including energy, manufacturing, utilities, and transportation. In these environments, millions of assets and billions of data flows require real-time security. The combined solution allows for the governance and enforcement of access to assets and data at an unprecedented scale, proactively preventing unauthorized activities and containing potential threats before they can disrupt essential services.
Also Read:
- Check Point and NVIDIA Unveil AI Cloud Protect for Enterprise AI Security
- CrowdStrike and NVIDIA Unveil Autonomous AI Agents for Advanced Cyber Defense
Xage is showcasing its integration with Nvidia BlueField at the Nvidia GTC Conference in Washington, D.C., which runs through October 29.
