Databricks unveiled OpenSharing on August 26, 2026, as the successor to its Delta Sharing protocol introduced in 2021 [1, 2]. OpenSharing is an open, vendor-neutral project hosted by the Linux Foundation that allows organizations to securely share AI assets including agent skills, AI models, unstructured data, and structured tables [1, 2].

The platform adds support for Apache Iceberg REST Catalog clients and integrates with on-premises storage vendors like Everpure, MinIO, and Qumulo, broadening its compatibility across systems [1, 2]. It supports zero-copy secure sharing methods which avoid duplicating or moving data and AI assets [1, 2]. APIs enable discovery, authorization, and access to AI assets without depending on proprietary platforms [2].

Delta Sharing, launched by Databricks in 2021, was the industry's first open protocol for secure data sharing between organizations and remains widely adopted by thousands of customers [1, 2]. Databricks co-founder and CTO Matei Zaharia said, "Delta Sharing proved the industry would choose open over locked-in" [1].

Akram Chetibi, product lead at Databricks, said OpenSharing addresses earlier challenges of sharing AI assets: "There is no easy way today to share an agent skill... You would email me a file, but then if I want an update, how do I get the update? You email me another file." He emphasized OpenSharing provides an open protocol allowing organizations to publish and consume AI assets regardless of platform [1, 2].

Industry experts see OpenSharing as a significant step for AI infrastructure. Analyst Stephen Catanzano of Omdia called it a "solid advance in AI infrastructure," noting its importance in extending secure zero-copy sharing from structured data to agent skills and AI models, assets increasingly critical in the current autonomous AI era [2].

OpenSharing was open sourced on GitHub and began Linux Foundation hosting on August 10, 2026, ahead of its launch by Databricks later in the month [2]. It supports cross-platform and cross-department sharing of AI models, agent skills, workflows, and unstructured data [2].

The next step for OpenSharing is broader adoption by industry players leveraging its APIs and open protocols to enable seamless, secure AI asset sharing at scale.