VAST Data launches DataEnclave to securely connect AI models with sensitive enterprise data

VAST Data has launched DataEnclave, a confidential AI capability designed to allow organisations to run leading artificial intelligence models against sensitive enterprise data while protecting both customer information and proprietary model technology.

The capability is part of the VAST AI Operating System and is built on NVIDIA Confidential Computing, with support from AI model developers, cloud providers, cybersecurity companies and hardware partners.

DataEnclave is designed for organisations in sectors such as financial services, healthcare and government, where sensitive information is often stored in tightly controlled environments and cannot easily be transferred to external AI services.

Secure AI processing inside trusted environments

According to VAST Data, DataEnclave enables proprietary and open AI models to be deployed within customer data centres or trusted cloud hardware.

The system uses a hardware-isolated secure runtime and cryptographic attestation to verify the computing environment and applicable security policies before sensitive data and AI models are decrypted and loaded into the protected environment.

Once processing begins, sensitive data and model assets remain protected within CPU and GPU memory.

Customer data encryption keys remain under the customer's control, while model keys and weights remain within the model builder's trust domain. VAST Data said infrastructure operators and administrators cannot access either during processing.

AI models for sensitive enterprise data

The new capability is supported by an ecosystem of AI companies and technology providers, including Cohere, CrowdStrike, Deepgram, Factory, Fundamental, NVIDIA and TwelveLabs, among others.

The approach is intended to address a challenge faced by organisations handling sensitive information: data may not be permitted to leave a controlled environment, while AI model developers may be unwilling or unable to deploy proprietary models into infrastructure they cannot trust.

DataEnclave is designed to provide a secure environment where the two can operate together without requiring organisations to relinquish control over their data or model providers to expose their proprietary technology.

Models treated as enterprise resources

VAST Data said the launch forms part of its broader vision for an AI operating system in which AI models are managed as resources alongside enterprise data.

As organisations increasingly use multiple specialised AI models for different applications, businesses will need to determine which model is suitable for a particular task, what information it can access, who can use it and which policies govern its operation.

The company also highlighted the growing importance of protecting model weights as organisations develop customised models and AI agents. Such models can represent valuable intellectual property and may require security controls comparable to those applied to other critical enterprise assets.

“Models are becoming a resource the operating system has to manage, the same way it manages data,” said Renen Hallak, Founder and CEO of VAST Data.

He said the company's approach is aimed at securely bringing advanced AI models to sensitive enterprise data while allowing organisations to manage model access, security and governance within existing operational boundaries.

The launch comes as businesses in highly regulated industries increasingly explore AI applications while facing requirements around data privacy, security and control over proprietary information.