TechPrufer
Get started
Coreweave
Coreweave logoCoreweave · Feature trail

AI Infrastructure Security

4 posts · May 2026 – Sep 2026

What this trail covers

CoreWeave enhances its AI infrastructure security by integrating NVIDIA BlueField-4 DPUs for hardware-enforced tenant isolation, offloading security functions, and extending zero-trust principles. This provides enhanced performance, scale, security, and efficiency for AI workloads, with telemetry flowing into existing SIEMs and automated enforcement of security states. This post details the integration of federated identity management (SAML/OIDC) via CoreWeave IAM with existing IdPs (e.g., Microsoft Entra, Okta) for automated user provisioning (AUP) and authorization. It also introduces Remote Key Encryption (RKE) for encrypting data at rest using keys managed entirely within customer-controlled key stores (e.g., HashiCorp Vault, KMS, HSM), ensuring CoreWeave only handles ciphertext. RKE supports automated key lifecycle management (rotation, expiration, revocation) and integrates with existing KMS/HSM infrastructure.

All Coreweave trails
  1. Enterprise AI Security on CoreWeave: IAM and Key ControlThis post introduces two key security enhancements for AI workloads on CoreWeave: federated identity management and remote key encryption. CoreWeave IAM federates with existing enterprise Identity Providers (IdPs) via SAML and OIDC, enabling Automated User Provisioning (AUP) for continuous user and group synchronization. This extends to SUNK User Provisioning (SUP) for Slurm-on-Kubernetes clusters, automatically creating POSIX users, groups, and Slurm accounts. Remote Key Encryption (RKE) allows data at rest to be encrypted using keys managed externally by the customer's KMS or HSM. Encryption occurs client-side within the customer's trusted compute boundary, and CoreWeave only handles ciphertext. RKE supports automated key lifecycle management (rotation, expiration, revocation) by integrating with existing KMS/HSM infrastructure.
  2. CISOs' New Bar for AI Infra Security | CoreWeave BlogThis post details the new security bar for AI infrastructure, emphasizing automated enforcement, continuous visibility into existing SIEMs, hardware-enforced process isolation via NVIDIA BlueField-3 DPUs, and evidence-based trust. It outlines how these principles address machine-speed exploitation and the evolving shared responsibility model in AI security, posing specific questions for CISOs to evaluate providers.
  3. Powering AI Innovation: CoreWeave + NVIDIA BlueField-4 | CoreWeave BlogThis post details the integration of NVIDIA BlueField-4 DPUs into CoreWeave's servers, leveraging the NVIDIA DOCA software framework to push core networking and security functions onto the DPUs. This enables hardware-enforced tenant isolation using EVPN VXLAN for secure, separate virtual networks per customer, eliminating software bottlenecks and noisy neighbors. The BlueField-4 DPUs provide a 6x performance boost and 800 Gb/s throughput, enhancing data access, reducing inference latency, and strengthening tenant isolation in multi-tenant AI clusters. It also introduces the BlueField Advanced Secure Trusted Resource Architecture for finer-grained, hardware-based isolation.
  4. Security Engineered by DesignThis post details how CoreWeave integrates NVIDIA BlueField-3 DPUs into its AI infrastructure to enforce tenant isolation and strengthen security boundaries. It also discusses the use of immutable logs for traceability, expansion of encryption enforcement, and the achievement of SOC 2 Type II certification. The post highlights federated IAM, SCIM automation, and workload federation using OIDC for fine-grained access control, and the development of a next-generation framework for full-stack integrity using hardware-based attestation.