VAST Data and NVIDIA Launch DataEnclave for Confidential Enterprise AI

VAST Data has announced DataEnclave, a confidential AI capability of its VAST DataEngine that lets enterprises run leading proprietary and open AI models directly on their most sensitive data, inside their own data centers or trusted cloud hardware. The product was built in collaboration with NVIDIA and an initial group of model partners including Cohere, CrowdStrike, and Deepgram.

The problem it targets is structural. By McKinsey's count, more than 80 percent of global organizations have deployed generative or agentic AI in at least one function, yet in highly regulated sectors, the most valuable data often cannot legally or practically move to where frontier models run. In European banking, fewer than 30 percent of institutions have integrated generative and agentic AI into their model risk management frameworks, per McKinsey's EMEA survey.

'Enterprise data is essential to accurate, usable AI, and keeping business data confidential is critical to protecting IP in the age of agents,' said Justin Boitano, Vice President of Enterprise AI at NVIDIA. 'VAST Data's integration of NVIDIA Confidential Computing delivers protection for both enterprises and model builders, providing security, identity, permissions, governance and compliance as a foundation of the agent architecture.'

Technically, DataEnclave extends the VAST AI OS with CPU and GPU Trusted Execution Environments built on NVIDIA Confidential Computing. Cryptographic attestation verifies the environment before sensitive data and model weights are decrypted in memory, including across encrypted NVLink connections. Customers keep control of their keys through KMS integrations, model builders retain their weights, and infrastructure administrators can access neither.

The capability list includes hardware-isolated execution inside confidential VMs, verify-before-decrypt attestation, support for both connected and air-gapped deployments using open CNCF and Fortanix attestation, a tamper-proof audit trail covering attestation and key lifecycle events, and secure agent sandboxes that give autonomous AI agents isolated, policy-enforced execution environments.

Partner reaction highlights sovereignty demand. 'Customers around the world have unique regulatory and sovereignty requirements, and they are asking for AI that is encrypted end-to-end, not just at rest but in motion and during inference,' said Frank O'Dowd, Chief Revenue and Commercial Officer at Cohere. VAST founder and CEO Renen Hallak frames the bigger picture: models are becoming a resource the operating system must manage like data, and the end state is every organization governing an ecosystem of fine-tuned models that represent its true intellectual property.

DataEnclave is in preview today, with general availability scheduled for the first quarter of 2027.