Databricks has introduced OpenSharing, an open standard designed to let organizations securely share data and AI assets across platforms, clouds, and internal systems. The company announced the project at its Data + AI Summit, describing it as the next step beyond Delta Sharing, its earlier protocol for cross-platform data exchange.
OpenSharing is hosted by the Linux Foundation and is intended to support a broader set of AI-era assets than Delta Sharing did. Databricks said the protocol can be used to share agent skills, AI models, and unstructured data, in addition to structured datasets. The project is also available on GitHub.
The company framed the launch as an effort to reduce the custom integration work that enterprises often need when moving AI assets between vendors or business partners. Databricks said OpenSharing provides standard APIs for discovery, authorization, and access, allowing providers to publish assets once and make them available to multiple recipients without copying the underlying data.
Delta Sharing, which Databricks launched in 2021, became known as an open protocol for secure data collaboration. The company said thousands of customers and partners now use it, including large enterprises such as Amadeus, Atlassian, LSEG, SAP, Stripe, and The Trade Desk. With OpenSharing, Databricks is widening that model to cover assets used by AI agents and modern machine learning workflows.
A key addition is support for Apache Iceberg clients, which Databricks said broadens the number of tools that can receive shared assets. The company said this move extends the ecosystem beyond Databricks and other platforms already connected through Delta Sharing.
OpenSharing also adds a path for organizations that keep data on-premises or in private cloud environments. Through integrations with storage partners including Everpure, MinIO, and Qumulo, customers can connect those assets to cloud AI and analytics platforms without moving the data itself. Databricks said more storage partners are expected to join, including Cohesity, Commvault, HPE, NetApp, Nutanix, Rubrik, and VAST Data.
Databricks pointed to support from a range of companies across travel, finance, analytics, and enterprise data infrastructure. OpenAI said it supports open AI ecosystems and wants standard, secure ways to discover and authorize access to AI assets. Amadeus said the technology could help power trusted AI use in the travel industry. Atlassian said it is using OpenSharing in its data shares to provide customers with broader access to cloud data. LSEG said the protocol aligns with its strategy to let customers use data across tools, clouds, and models.
Other partners highlighted potential uses for private data environments and enterprise collaboration. MinIO said native support in its AIStor platform could help unlock data that cannot be moved. SAP said OpenSharing supports its plan for open collaboration around AI-first data products. Stripe said the protocol can help users work with customer, billing, and transaction data in their preferred environments. Acxiom said the standard can make data interoperable across identifiers, platforms, partners, and clouds.
Databricks said the main benefit of OpenSharing is that it offers a vendor-neutral way to share assets that are increasingly central to AI applications. As organizations deploy more agents and model-based services, the company argued, they need a standard approach that works across ecosystems rather than one tied to a single platform.
The launch comes as Databricks continues to expand its AI product line, while also pushing open infrastructure as a strategic alternative to closed marketplaces and proprietary sharing systems. OpenSharing is now available for developers and organizations to explore through GitHub and Databricks' platform.