VAST Data Launches Secure AI Platform for Sensitive Data
News related to:VAST Data · 3 min read
Remote-First-Company VAST Data, the AI Operating System company, today announced the launch of VAST DataEnclave, a confidential AI capability of the VAST DataEngine, built on NVIDIA Confidential Computing, in collaboration with ecosystem partners including top AI model builders, AI clouds, AI security, and leading AI hardware providers. The new offering extends the VAST AI Operating System to resolve the challenge of deploying proprietary and open AI models in environments where they could not previously operate, such as inside customer data centers or trusted cloud hardware, including environments where sensitive data remains tightly controlled.
VAST DataEnclave enables the execution of proprietary and open AI models in a hardware-isolated secure runtime, using NVIDIA Confidential Computing to encrypt guest memory, GPU memory, and NVLink traffic while isolating sensitive assets, such as proprietary models and sensitive data, from infrastructure operators, administrators, and other tenants sharing the same hardware. The secure container runtime and attestation service are directly integrated within the VAST DataEngine, ensuring that sensitive assets are only released during execution after the workload is verified and a secure enclave has been established.
Key capabilities of VAST DataEnclave include hardware-isolated execution, which protects workloads inside confidential virtual machines and containers, and verify-before-decrypt attestation, which cryptographically verifies the trusted execution environment, including NVIDIA GPU attestation, before releasing decryption keys. This ensures that sensitive assets, such as proprietary models and sensitive data, are accessible only to approved workloads running in a trusted environment.
Independent key control allows enterprises and model builders to maintain their respective keys within their own trust domains through Bring Your Own Key Management System (KMS) integrations, protecting enterprise's own fine-tuned weights, which are fast becoming critical intellectual property, as much as a model builder's base weights. VAST DataEnclave supports connected or fully air-gapped environments, with DataEnclave deployments using attestation services built on the open CNCF Trustee stack or in partnership with Fortanix via its Confidential AI infrastructure for fully sovereign AI.
The new offering also includes a governed and auditable environment, where attestation events, key releases, and enclave lifecycle actions are recorded in a tamper-proof, queryable audit trail in the VAST DataBase, providing visibility into what ran, where, and under what verified policy without exposing protected data or weights. VAST DataEnclave also provides secure agent sandboxes, which are isolated execution environments for AI agents, enforcing policy over the data, systems, and tools agents can access and the actions they can take.
VAST DataEnclave extends confidential execution across the infrastructure where AI models and sensitive data are processed, combining hardware isolation and verifiable attestation to protect both while they are in use. The announcement expands the range of advanced models organizations can deploy within their own environments, while advancing VAST's broader AI Operating System vision, in which models are managed as a logical resource alongside data rather than as applications that simply sit on top of the infrastructure.
VAST DataEnclave is expected to ship in the first quarter of 2027 through VAST Data and its ecosystem partners. The new offering is designed to bring leading AI models securely to the world's most sensitive data, enabling organizations to fine-tune their own models and manage specialized models across environments, determining where they run, what data they can access, who or what can use them, and the policies under which they operate. The announcement also highlights the need for secure management at scale, as organizations increasingly fine-tune their own models and generate specialized intelligence from their interactions, with model weights becoming a new class of enterprise intellectual property.
VAST Data, in collaboration with NVIDIA and partners, is bringing proprietary and open AI models across a range of modalities to the world's most sensitive data, while giving customers control over cost, performance, model selection, and data privacy. The new offering is expected to expand the range of advanced models organizations can deploy within their own environments, while advancing VAST's broader AI Operating System vision, in which models are managed as a logical resource alongside data rather than as applications that simply sit on top of the infrastructure.
VAST DataEnclave extends confidential execution across the infrastructure where AI models and sensitive data are processed, combining hardware isolation and verifiable attestation to protect both while they are in use. The announcement expands the range of advanced models organizations can deploy within their own environments, while advancing VAST's broader AI Operating System vision, in which models are managed as a logical resource alongside data rather than as applications that simply sit on top of the infrastructure. The new offering is expected to ship in the first quarter of 2027 through VAST Data and its ecosystem partners.