Databricks Contributes OpenSharing Protocol to Linux Foundation to Standardize AI Asset Exchange

Evolving from Delta Sharing

OpenSharing builds directly on the widely adopted Delta Sharing protocol, also created by Databricks . Delta Sharing already enabled organizations to securely share live data from Delta Lake and other sources without copying it. OpenSharing extends this foundation to a broader class of AI assets, making it a natural upgrade path for the many enterprises already invested in the Delta ecosystem.

New Capabilities

OpenSharing introduces several important capabilities beyond the original Delta Sharing specification:

• Apache Iceberg Client Support: The protocol now supports Apache Iceberg IRC clients, significantly broadening cross-platform interoperability and making it easier for organizations that have standardized on Iceberg to participate in shared AI workflows .
• On‑Premises Storage Integrations: Initial integrations with on-premises storage partners MinIO and Qumulo allow organizations to keep data within their own infrastructure while still participating in the open sharing ecosystem. The Linux Foundation and Databricks have indicated that additional storage partners are expected to join soon .
• Unified Exchange Framework: Rather than treating AI models, data, and agent skills as separate problems, OpenSharing provides a single framework for exchanging all three asset types, reducing integration complexity .

Who Is Backing the Project?

The project has attracted a broad coalition of enterprise and AI leaders. Backing organizations include Amadeus, Atlassian, LSEG, SAP, Stripe, and OpenAI, among others . This mix of established enterprise software companies, financial infrastructure providers, and frontier AI labs suggests that OpenSharing is addressing a genuinely cross-industry need.

Where Is OpenSharing Available Now?

OpenSharing is hosted at opensharing.io under the governance of the Linux Foundation . This neutral governance structure is critical for fostering trust among competitors and ensuring the protocol evolves in the interests of the broader community rather than a single vendor.

The Bigger Picture

OpenSharing arrives at a moment when the AI industry is grappling with fragmentation at multiple layers—from agent communication protocols to model licensing. The Linux Foundation has been active on several related fronts, including earlier work on the Agentic AI Foundation (AAIF), which provides a neutral home for projects like Anthropic’s Model Context Protocol (MCP), and the Agent2Agent (A2A) protocol for agent communication. OpenSharing complements these efforts by focusing specifically on the data and asset layer that powers AI workloads.

The protocol’s emphasis on on-premises storage support also signals an understanding that many regulated industries will need to keep sensitive data within their own boundaries even as they collaborate on AI. By securing commitments from MinIO, Qumulo, and the promise of more on-premises partners, the project is positioning itself for adoption in finance, healthcare, and other sectors with strict data sovereignty requirements.