The Announcement
The Linux Foundation has launched the OpenSharing Project, an open, vendor-neutral protocol for sharing AI assets and data across organizations and platforms. The project was contributed by Databricks and builds directly on the Delta Sharing protocol, extending its scope well beyond structured data to include AI models, agent skills, and unstructured data. The goal is to create a common interoperability layer for the emerging agentic AI ecosystem, with support for Iceberg recipients and other open table formats. In plain terms: enterprises that want their AI agents, models, and data to move across systems today largely have to build custom integrations or accept proprietary marketplace dependencies. OpenSharing is an attempt to change that.
The Bigger Picture
The Interoperability Problem Is Real and Getting Worse
Enterprise AI adoption has accelerated sharply. According to ECI Research, 92% of organizations report that AI capabilities are now integrated into at least one stage of their software delivery lifecycle, up from 71% in early 2024. That pace of adoption has outrun the infrastructure needed to support it. Most enterprises are not running AI in a single, coherent environment. They are running it across a patchwork of tools, teams, and platforms, and the connective tissue between those systems is overwhelmingly bespoke or proprietary.
The Delta Sharing protocol, which OpenSharing extends, solved a narrower version of this problem for structured data. It worked because it was simple, open, and adopted quickly across a broad vendor ecosystem. Databricks is betting that the same formula applies to AI assets, and contributing this to the Linux Foundation is a deliberate signal: this is intended to be infrastructure, not a product feature.
That framing matters. The Linux Foundation’s track record in stewarding foundational internet and cloud standards gives the OpenSharing Project credibility that a vendor-controlled initiative would not have. The inclusion of Iceberg recipient support also aligns OpenSharing with the open table format ecosystem that is increasingly becoming the default for enterprise data architecture, which reduces the integration friction enterprises would otherwise face.
What This Means for IT Decision-Makers
The business case here is about reducing the cost and risk of AI fragmentation. Today, sharing a model or an agent capability across organizational or platform boundaries typically requires either negotiating access to a proprietary marketplace or building a custom integration that becomes a maintenance liability. Neither option scales well, and both create vendor concentration risk.
OpenSharing introduces a third option: an open standard that any platform can implement. For ITDMs, the near-term implication is not immediate deployment. It is vendor evaluation. Organizations that are currently selecting AI platforms, data platforms, or agentic AI infrastructure should treat OpenSharing protocol support as a forward-looking evaluation criterion. Platforms that adopt it early will be meaningfully easier to integrate with over a three-to-five year horizon. Platforms that resist adoption are, in effect, choosing a walled garden strategy.
ECI Research found that two-thirds of enterprise AI leaders have already implemented multi-agent collaboration in live or pilot workflows. That figure signals how quickly multi-agent architectures are becoming operational, not experimental. As agent workflows span more systems and organizational boundaries, the absence of a common protocol for exchanging agent skills and model artifacts becomes an active drag on deployment velocity.
What This Means for Developers
For developers building agentic AI systems, the OpenSharing Project could address a friction point that is familiar but rarely acknowledged in vendor literature: the handoff problem. Moving a fine-tuned model or a reusable agent skill from one environment to another currently involves a mix of serialization formats, access control conventions, and integration code that varies by platform. There is no HTTP-equivalent for AI asset exchange. OpenSharing is an attempt to create one.
The extension to unstructured data is particularly relevant. Production agentic workflows rarely operate on clean, structured datasets. They consume documents, logs, embeddings, and other unstructured artifacts. A protocol that handles only tabular data would leave a significant share of real-world agent workflows unaddressed. By explicitly scoping OpenSharing to include unstructured data and agent skills, the project is designed for the workflows developers are actually building, not the idealized ones.
The practical near-term question for developers is ecosystem adoption. A protocol is only as useful as the platforms that implement it. Delta Sharing’s adoption trajectory is encouraging here, but the AI asset exchange space has more incumbents with stronger lock-in incentives than the structured data sharing space did. Developers evaluating platforms should watch for OpenSharing SDK availability and native support as indicators of genuine vendor commitment.
What’s Next
Protocol Adoption Will Define the Timeline
OpenSharing’s impact will be determined by the speed and breadth of ecosystem adoption, not by the protocol specification itself. The Linux Foundation provides governance credibility, but adoption requires platform vendors, cloud providers, and framework maintainers to ship implementations. The Delta Sharing precedent is positive: broad adoption came relatively quickly once the protocol was simple and the Linux Foundation backstop was in place. The OpenSharing scope is larger, which means the adoption curve will likely be longer, but the foundational pattern is replicable.
Organizations should watch for cloud provider and major AI platform announcements over the next six to twelve months as the first real signal of adoption momentum. Early implementations from two or three major cloud or data platform vendors would be a meaningful inflection point.
Governance and Security Will Determine Enterprise Trust
Open protocol or not, enterprises will not exchange AI models and agent skills across organizational boundaries without confidence in the governance and security model. The OpenSharing Project will need to address credential management, provenance attestation, and access control with the same clarity that Delta Sharing brought to data access. ECI Research found that nearly half of respondents rate compliance and data governance as a high priority when developing AI systems. That concern intensifies when assets cross organizational lines.
The near-term roadmap visibility on governance capabilities will be as important as the protocol specification itself. For enterprise adopters, the question is not whether open interoperability is desirable in principle. It clearly is. The question is whether the governance model is mature enough to meet the compliance requirements of regulated industries, which are precisely the organizations with the most to gain from frictionless AI asset exchange.
Stay Ahead of Application Development Trends
Get weekly analyst insights, research notes, event coverage, and AppDevANGLE updates delivered directly to your inbox.
Subscribe for Weekly Insights
Join technology leaders, practitioners, and GTM teams following the trends shaping modern software delivery.
Looking for deeper research access?
Explore ECI Research reports, survey insights, and market analysis through the ECI Research Portal.
