Databricks has unveiled OpenSharing, an open protocol under the Linux Foundation AI & Data Foundation that enables enterprises to share AI models, agent skills, dashboards, and unstructured data across platforms without copying or moving assets. The protocol uses a zero-copy credential vending model, issuing temporary scoped credentials so recipients access assets directly from the provider's cloud storage. Analysts highlight that it addresses the growing 'integration tax' of enterprise AI — the overhead of packaging, translating, and syncing AI assets across organizational boundaries. Unlike Snowflake's zero-copy approach, which requires both parties to be on Snowflake, OpenSharing works cross-platform. It evolves from Databricks' Delta Sharing protocol and currently supports connectors for Python, Spark, Tableau, PowerBI, Snowflake, DuckDB, Java, Rust, Go, and more.

4m read timeFrom infoworld.com
Post cover image
Table of contents
Reducing the integration tax of enterprise AIFrom data sharing to AI asset sharing
42 Impressions