The Linux Foundation launches OpenSharing Project, an open, vendor-neutral protocol that standardizes sharing of AI assets and data across platforms, extending Delta Sharing to include agent skills, AI models, and unstructured data volumes.

OpenSharing builds on Delta Sharing connectors to broaden interoperability to Iceberg IRC clients, increasing cross-platform reach and reducing data-sharing fragmentation across ecosystems.

Spokespersons stress the need for open, interoperable AI asset exchange to accelerate innovation and minimize ecosystem fragmentation.

Databricks introduces the Storage (SDS) Ecosystem, a partner category enabling the Databricks Intelligence Platform to access on-premises, private cloud, or edge data without moving it.

A unified catalog across hybrid environments enables querying and model training on on-prem data without data leaving premises; customers can use Serverless Compute, Genie, and AgentBricks with OpenSharing.

The project is available on GitHub and inside the Databricks platform, with additional partners and storage integrations planned, including Cohesity, Commvault, HPE, NetApp, Nutanix, Rubrik, and VAST Data.

OpenSharing enables on-premises and private-cloud data sharing through native integrations with Everpure, MinIO, and Qumulo, allowing direct cloud connections without data movement.

As an open-source protocol, OpenSharing secures governed data sharing by exposing on-premises data estates to Databricks Serverless Compute via Unity Catalog, enabling zero data movement and zero duplication.

Launch partners MinIO (AIStor), Everpure, Qumulo, and Vast Data integrate OpenSharing to connect on-prem data with Databricks, emphasizing secure, governed access without data replication.

The overarching message is that enterprises can govern and analyze data across cloud, on-prem, and edge without moving data, enabling hybrid AI and scalable, compliant data estates.

A next focus is on unlocking unstructured data via Volumes APIs to expose on-premises unstructured files (images, PDFs, videos, scans) to GenAI workloads, expanding beyond structured data.

Databricks shifts to a governance-centric approach, driven by data sovereignty, cost, latency, and AI-on-dark-data considerations.