The project, contributed by Databricks and now hosted under Linux Foundation governance, extends the Delta Sharing protocol beyond structured data into AI-era assets such as agent skills, machine-learning models and unstructured data volumes. Its backers say the goal is to reduce reliance on proprietary marketplaces, one-off integrations and duplicated data pipelines that have slowed cross-platform AI deployment.
OpenSharing arrives as companies seek to build AI systems that can operate across clouds, data lakes, private infrastructure and business partners without locking sensitive information into a single vendor stack. The protocol is designed to let organisations publish data and AI assets once, with standard mechanisms for discovery, authorisation and access, while allowing recipients to consume them through different platforms.
The Linux Foundation said the project is intended to provide community-governed infrastructure for secure AI collaboration at scale. Jim Zemlin, its chief executive, said OpenSharing addresses “a critical need” for a common framework that lets organisations exchange AI assets securely and interoperably across platforms and ecosystems.
Databricks has positioned the move as the next stage of Delta Sharing, which it introduced in 2021 for secure data exchange. Delta Sharing has been used across analytics and business intelligence tools including Apache Spark, Oracle, Power BI, Tableau and Snowflake, and by enterprises including Amadeus, Atlassian, LSEG, SAP, Stripe and The Trade Desk. OpenSharing broadens that base by adding support for Apache Iceberg recipients and by covering AI assets that were not part of traditional data-sharing workflows.
Matei Zaharia, Databricks co-founder and chief technology officer, said Delta Sharing had shown that the industry would choose “open over locked-in”. He said OpenSharing extends that principle to the full AI stack, including Iceberg recipients and on-premises providers.
A central selling point is the ability to connect on-premises and private-cloud data sources to cloud-based AI and analytics systems without moving the underlying data. Storage partners including Everpure, MinIO and Qumulo are listed among early providers, with Cohesity, Commvault, HPE, NetApp, Nutanix, Rubrik and VAST Data named as additional partners. That architecture could be significant for financial services, healthcare, telecoms and public-sector users that face data residency, audit and security requirements.
The announcement reflects a wider shift in enterprise AI from model selection towards infrastructure control. Companies deploying autonomous or semi-autonomous agents increasingly need standard ways for systems to discover capabilities, use data, call tools and exchange outputs across organisational boundaries. Without common protocols, agents can remain trapped in isolated software environments, limiting their usefulness in procurement, compliance, customer operations, logistics and software development.
OpenSharing is not the only standardisation effort gaining ground. The Agent-to-Agent protocol, originally developed by Google and now hosted by the Linux Foundation, focuses on how AI agents discover, communicate and transact across frameworks and vendors. The Model Context Protocol has also become an important part of the integration debate by standardising how agents connect to tools and enterprise systems. OpenSharing’s narrower role is to address the exchange of data and AI assets rather than agent communication alone.
The launch also comes amid evidence that AI adoption is creating as many organisational bottlenecks as technical breakthroughs. A European technology talent study found employers expect AI to have a positive net hiring effect of 27% in 2026 and 17% in 2027, with demand particularly strong for AI-specific roles. Yet security concerns affect 51% of organisations and privacy concerns 44%, while gaps in cybersecurity, AI operations, cloud computing and data engineering are limiting deployment.
Europe’s challenge is especially acute. AI talent in the EU more than doubled between 2016 and 2023, but still represents only 0.41% of the workforce. Women account for less than a quarter of AI engineers across Europe, and in some cities the share is as low as 11%. Germany issued nearly 78% of all EU Blue Cards in 2023, highlighting how skilled migration remains concentrated in a few markets rather than evenly spread across the bloc.
Enterprises are responding by prioritising internal training. Upskilling existing staff is now favoured over external hiring because it preserves institutional knowledge, lowers recruitment costs and improves team cohesion. Open source has also become a core strategy for AI implementation, particularly for organisations seeking sovereignty, lower licensing costs and reduced dependence on dominant technology suppliers.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
