New Linux Foundation-hosted standard enables secure sharing of AI models, agent skills and enterprise data across cloud, on-premises and multi-platform environments.
A new open-source protocol OpenSharing by Databricks aimed at simplifying how organizations exchange data and AI assets has been launched, extending the capabilities of the widely adopted Delta Sharing framework into the AI era.
Hosted by the Linux Foundation, the initiative introduces a vendor-neutral approach for sharing AI models, agent skills and unstructured data across organizations. The technology is designed to eliminate the need for custom integrations and proprietary marketplaces that have traditionally complicated AI collaboration between enterprises.The release marks the next phase of Delta Sharing, the open protocol introduced in 2021 that has become a common method for secure, zero-copy data exchange. Thousands of organizations use the framework today, including major enterprises such as Amadeus, Atlassian, LSEG, SAP and Stripe.
One of the most significant additions is support for Apache Iceberg APIs. The capability allows data providers to publish assets once and distribute them to a broader ecosystem of analytics, AI and data-management tools, increasing interoperability while reducing platform lock-in.Beyond structured datasets, enterprises can now securely expose AI capabilities directly from the source. Agent skills and machine-learning models can be discovered, authorized and accessed through standardized interfaces, enabling partners and customers to consume these resources without moving or duplicating underlying assets.
The platform also addresses a growing challenge for organizations operating under strict data residency and compliance requirements. Through integrations with storage providers such as Everpure, MinIO and Qumulo, cloud-based analytics and AI services can access information stored on-premises or in private-cloud environments without requiring data migration.Additional support is expected from infrastructure vendors including Cohesity, Commvault, HPE, NetApp, Nutanix, Rubrik and VAST Data, expanding deployment options for enterprises with hybrid environments.
The effort has received backing from several technology and enterprise software companies. OpenAI, SAP, Atlassian, LSEG and Stripe highlighted the importance of open standards for enabling secure collaboration, data portability and broader AI adoption across industries.The GitHub release is available immediately, providing developers and enterprises with a common foundation for sharing data and AI assets across diverse platforms while maintaining control over where those assets reside.
