Databricks introduced OpenSharing, an open-source protocol hosted by the Linux Foundation, on June 10 to standardize how organizations share AI assets. The new framework extends the company's existing Delta Sharing protocol to include AI models, agent skills, and unstructured data, allowing teams to collaborate across different platforms without duplicating files.
Extending data sharing to AI assets
Five years ago, Databricks built Delta Sharing to let enterprises share structured data securely. OpenSharing builds on that architecture to address a gap in how teams handle AI collaboration.
When organizations attempt to share AI models, prompts, or specialized agent workflows across company boundaries, they currently must build custom solutions for every new partnership. OpenSharing provides a single, open protocol for publishing and consuming these assets regardless of the underlying platform.
The protocol also adds support for the Apache Iceberg REST Catalog, accommodating organizations that use Iceberg-native tools. Furthermore, it integrates with on-premises platforms and private clouds, ensuring enterprises can share assets without moving data out of secure environments.
Industry response to standardized AI collaboration
Analysts point out that autonomous agents are becoming primary consumers of data, requiring new methods to manage distributed assets.
"OpenSharing marks a shift from simple data exchange to a unified, governed interface for the AI and data stack," William McKnight, president of McKnight Consulting, said. "Beyond traditional tables … this framework provides a blueprint for studying and scaling how autonomous agents interact with distributed data. This could be quite significant for data sharing."
Stephen Catanzano, an analyst at Omdia, said the protocol addresses the fragmented environments where AI assets currently reside.
"What makes it particularly important is that it extends secure, zero-copy sharing beyond structured data to include agent skills and AI models -- assets that are becoming critical in the agentic era," Catanzano said. "Previously, organizations had no standardized way to share these AI components across platforms."
Broader push for open AI standards
OpenSharing is part of a wider industry movement to create vendor-neutral standards for AI development and deployment. For example, the Model Context Protocol (MCP), released by Anthropic in late 2024, standardizes how AI models connect to proprietary data sources.
Similarly, the LF AI & Data Foundation recently formed the DocLang Specification Working Group. This group aims to develop an open, AI-native document format to standardize how document data is prepared and governed for AI systems, addressing bottlenecks in processing unstructured enterprise knowledge like PDFs and slides.
Akram Chetibi, director of product management at Databricks, said making these protocols open source prevents vendor lock-in and supports flexible AI for Product Development workflows.
"Customers and partners don't want to collaborate with their data and AI assets while being confined to a single vendor proprietary ecosystem," Chetibi said. "It simply doesn't stick because it constrains innovation."
Why this matters for product development
Product development teams building agentic AI systems often struggle with fragmented tooling when collaborating with external partners or cross-functional departments. OpenSharing and similar open protocols remove the need to rebuild data pipelines or copy files manually for every new integration.
By adopting these standards, product teams can treat AI models, agent skills, and metadata as governed, shareable packages. This reduces development lag and allows engineers to focus on building capabilities rather than maintaining custom sharing infrastructure.
Your membership also unlocks:
