Key Takeaways
- VAST Data unveiled its agentic AI platform at the VAST Forward 2026 event, enhancing its AI Operating System.
- The platform features two new components: PolicyEngine, which governs AI access to data, and TuningEngine, which supports continuous model learning.
- PolicyEngine applies rules, logs activities, and monitors AI agents to ensure secure operations and prevent data leakage.
- TuningEngine uses techniques like LoRA fine-tuning and reinforcement learning to improve AI models through ongoing feedback and testing.
- Both engines integrate within the DataEngine framework, aiming for deployment by the end of 2026, facilitating self-learning AI systems.
The VAST Data agentic AI platform was unveiled to support secure, trusted, and self-learning AI systems. The announcement took place at the VAST Forward 2026 event. The platform expands the company’s AI Operating System. It is designed to help enterprises deploy and manage agent-based artificial intelligence at scale.
Two new components form the core of the update. They are called PolicyEngine and TuningEngine. Both operate within the VAST AI Operating System.
PolicyEngine Adds Governance Controls
PolicyEngine manages how AI agents access data and external tools. It applies detailed rules to regulate agent behavior. The system logs activities and keeps audit trails. This allows monitoring of decisions made by AI agents.
PolicyEngine also controls interactions with shared memory and knowledge resources. It is built to reduce risks related to unauthorized access and data leakage. All actions performed by agents can be reviewed within the platform.
TuningEngine Enables Continuous Model Learning
TuningEngine supports ongoing improvement of AI models. It processes feedback generated during agent operations. The system uses techniques such as LoRA fine-tuning, supervised fine-tuning, and reinforcement learning.
The engine creates candidate models for testing. It benchmarks performance results. Updated versions can then be redeployed into the system. This process forms a continuous learning cycle.
Integrated Architecture for Agentic AI Systems
The VAST Data agentic AI platform integrates governance and automated tuning within a unified architecture. Both new engines operate inside the broader DataEngine framework. The system is designed for deployment across various computing environments.
The company stated that PolicyEngine and TuningEngine are expected to be available by the end of 2026. The platform aims to support AI systems that can operate, evaluate outcomes, and improve based on collected data.