VAST Data and TwelveLabs have announced a partnership aimed at expanding large-scale video intelligence beyond public cloud environments, targeting organisations managing vast and highly governed video archives.
Unveiled at VAST Forward 2026, the collaboration introduces what the companies describe as TwelveLabs’ first customer-managed deployment path on the VAST AI Operating System. The move is designed to support enterprises and government agencies that require on-premises or hybrid control over sensitive video data due to sovereignty, regulatory or security constraints.
Video data presents unique operational challenges. Its size, complexity and governance requirements often make it difficult to search, analyse and extract insight at scale. TwelveLabs develops video foundation models intended to enable multimodal understanding across visuals, speech, sound and temporal context. Its Marengo model focuses on multimodal embeddings and search, while Pegasus is designed for deeper video reasoning and text generation.
Under the partnership, these models will run within the VAST AI Operating System, which is built to manage unstructured data at exabyte scale. The integration allows organisations to deploy advanced video search, analytics and reasoning workflows closer to where the data is generated and stored, rather than relying solely on public cloud infrastructure.
The approach is positioned for environments where cloud-only deployment may be limited by data sovereignty requirements, cost concerns or internal security policies. By enabling customer-managed deployment, the companies are targeting sectors where video archives are both extensive and tightly controlled.
The VAST AI Operating System provides several foundational components intended to simplify large-scale AI workflows. These include a unified global namespace through VAST DataSpace to manage video across hybrid multicloud environments; built-in vector storage and retrieval via VAST DataBase, supporting similarity search at trillion-vector scale; and real-time orchestration through VAST DataEngine to generate and act on embeddings and metadata as video enters the system.
The partners see demand emerging across multiple industries. In media and entertainment, video intelligence can support faster archive search, content enrichment and production workflows. In financial services, it may be applied to surveillance footage for compliance and fraud detection. Public sector agencies are also exploring large-scale video analysis for investigations and situational awareness across multi-camera environments, particularly where infrastructure must remain under direct organisational control.
John Mao (pictured), Vice President of Global Technology Alliances at VAST Data, said the partnership is intended to allow organisations to bring AI capabilities directly to governed data environments rather than relocating sensitive content.
The announcement reflects a broader industry shift toward deploying advanced AI models within controlled or hybrid infrastructure environments, particularly as organisations grapple with the trade-offs between scalability, cost and regulatory compliance when managing large unstructured datasets such as video.
