What Databricks OpenSharing Is and Why It Matters
Databricks OpenSharing is an open AI model sharing protocol that defines common ways for organizations to publish, discover, authorize, and access AI models, agent skills, and data assets across different platforms and environments without being tied to a single vendor. Building on the earlier Delta Sharing standard, OpenSharing moves from data-only exchange to a broader view of how enterprises collaborate in the agentic era. Databricks describes OpenSharing as “the first open protocol to cover agent skills, AI models, and unstructured data,” extending the idea of open interoperability to the full AI stack. For data teams already managing lakehouses and mixed toolchains, this means a single, protocol-driven path to expose governed assets to partners, applications, and AI agents, while retaining control over authorization and lineage instead of building one-off integrations or relying on closed marketplaces.
From Delta Sharing to OpenSharing: Expanding the Interoperability Scope
Delta Sharing proved that open data protocols can scale, with thousands of customers and partners using it for secure collaboration, sharing, and monetization of datasets. OpenSharing is framed as the “next evolution” of this work and is now a project of the Linux Foundation, signaling a push for neutral governance beyond a single vendor. Technically, it broadens the unit of exchange from tables to a wider set of AI assets, including unstructured data and agent skills that live closer to production AI systems. The protocol also adds support for Iceberg IRC clients, which allows asset providers to reach a wider range of data recipients running different table formats and engines. For enterprises that already operate mixed storage and query stacks, this expansion turns Delta-era interoperability into a more complete cross-platform AI collaboration capability, rather than a lakehouse-only feature.
Secure Cross-Platform AI Collaboration Without Lock-In
Before OpenSharing, enterprises wanting cross-platform AI collaboration had to wire together custom APIs, build point-to-point integrations, or commit to a single marketplace controlled by one provider. OpenSharing addresses this by defining standard APIs for discovery, authorization, and access that any compatible platform can implement. This includes support for on-premises storage providers, so organizations can expose governed assets from local systems directly to cloud platforms without physically moving the data. According to Databricks, OpenSharing enables asset providers to reach Iceberg recipients and on-premises environments through one open protocol, rather than maintaining separate distribution paths. For security teams, the promise is consistency: policy can be expressed once at the protocol level and applied across different tools, instead of being reimplemented in each integration, helping reduce both operational risk and integration cost.
Implications for Enterprise Data Integration and AI Workflows
Enterprises already using the Databricks Platform show how a unified, governed foundation changes how AI shows up in day-to-day work. Applied Materials, for example, reports more than 1,500 analysts running 17 million self-service queries and over 100 machine learning models in production, with development 90 percent faster after moving to a lakehouse. Virgin Atlantic and Fonterra use Databricks to bring together operational, financial, and customer data so decisions can be made from shared metrics instead of isolated reports. In this context, Databricks OpenSharing extends that unified model outward: instead of keeping AI models and agent skills inside a single lakehouse, enterprises can treat them as shareable products in a wider ecosystem. This deepens enterprise data integration by turning internal AI capabilities into external, protocol-compliant services while preserving governance.
Databricks as Collaboration Infrastructure for the Agentic Era
OpenSharing also signals a strategic shift for Databricks from being viewed mainly as a data and AI platform to acting as an infrastructure layer for cross-platform AI collaboration. By donating the project to the Linux Foundation and extending support to Iceberg clients and on-premises providers, Databricks is positioning the protocol as a neutral standard rather than a proprietary extension. For data teams, this creates an architectural pattern: lakehouse for unified storage and governance, OpenSharing for outward AI and data exchange. As agent-based applications spread, the ability to publish a proprietary agent skill or AI model once and have partners consume it anywhere becomes a design requirement, not a nice-to-have. OpenSharing aims to be the connective tissue for that pattern, helping enterprises avoid new silos even as their AI landscapes grow more complex and distributed.
