CoreWeave Zero Trust AI Framework | CoreWeave Blog

This post introduces the CoreWeave Zero Trust AI Framework, detailing six layers of protection for AI workloads: Remote Key Encryption (RKE) for customer-controlled encryption keys, Support Access Management (SAM) for explicit customer authorization of support access, attestation services for verifying infrastructure integrity, and confidential computing (planned for NVIDIA Blackwell and Vera Rubin) to protect data in use. It emphasizes customer sovereignty over models, data, and runtime environments, allowing customers to bring their existing security systems to CoreWeave.