What’s New with Databricks Clean Rooms

Databricks Clean Rooms enables customers and partners to analyze their combined data, without revealing sensitive raw data to one another. Since the GA launch on AWS and Azure in February 2025, we've seen many customers adopt clean rooms in Advertising, Financial Services, Healthcare and many other industries. In this blog post we highlight how leading identity partners are leveraging clean rooms for privacy centric identity-resolution, as well as new collaboration and privacy functionalities that customers have asked for.

Identity Resolution in Clean Rooms

We’re excited to introduce an advancement in identity resolution: secure, cloud-native identity resolution in Databricks Clean Rooms. This innovation empowers partners and marketers to confidently and securely match and enrich datasets using common identifiers—all without exposing raw PII or moving data off-platform.

To enable this capability, we’re thrilled to partner with industry leaders including Epsilon, Deloitte, LiveRamp, and Acxiom. Together, we’re making it easier for organizations to unify fragmented records, connect related data, and unlock richer insights—all within a privacy-centric environment.

Consider an e-commerce brand collaborating with a major identity provider to resolve hashed customer emails, postal data, and device touchpoints, to a person or household based identity. With this new capability, both parties can bring data into a clean room and perform identity resolution in place—safely and efficiently.

This approach marks a new era in data collaboration: one where sensitive data never leaves the platform, identity matching happens seamlessly, and insights are generated without compromise. Databricks is proud to power this shift, offering a scalable environment that redefines what’s possible for clean-room collaboration and identity resolution.

Clean Rooms on Google Cloud Platform (GCP)

Clean Rooms is now generally available on Google Cloud, enabling seamless cross-cloud collaborations with full flexibility. Starting today, customers create a central clean room environment on GCP and collaborate with partners across AWS, Azure, or any other data platform. This aligns with our “Any cloud, any platform” philosophy: “Collaborate in a privacy-centric environment across clouds, regions and data platforms without requiring data movement”.

For example, a large retailer on GCP and a consumer brand on AWS, looking to partner and analyze the effectiveness of their joint marketing efforts. The retailer can now spin up a clean room on GCP and invite the brand to securely contribute their engagement metrics from recent campaign data. By combining this with the retailer’s customer purchase data, both parties can collaboratively identify trends and measure the impact of promotions—without ever sharing or exposing their raw data to each other.

Other clean room providers limit you to collaborate on their cloud or platform. With Databricks Clean Rooms, you avoid cloud silos and vendor lock-in and data stays in place via Delta Sharing, and you only share anonymized outputs.

Multi-Party Collaboration

Clean Rooms now support multiple collaborators in a single room. Previously, each clean room was effectively two-party only; now you can invite up to 9 other organizations (i.e., 10 total). These collaborators can be on different clouds, regions, or data platforms, yet work together in one central environment. This unlocks “Any scale, any trust level” and supports many-to-many collaborations with fine-grained access controls and orchestration.

Consider a retail marketing scenario: a retailer, its advertising brand, and a market research firm want to build a combined customer insights model. All three parties bring proprietary data and code. With multi-party clean rooms, the retailer can invite both partners into one clean room, share necessary tables, and run joint analytics. For instance, the retailer’s e-commerce data, the brand’s customer data, and the researcher’s survey data can be joined and analyzed together, without any party ever seeing each other’s raw tables.

By enabling multi-party collaboration with fine-grained governance, you unlock richer insights that require more than two organizations.

Flexible Privacy Approvals

Clean Rooms now support secure self-runs, allowing collaborators to upload and run their own notebooks with explicit approval from other clean room participants. Previously, notebooks could only be run by the other party, with approval implied by clicking the run button.

For example, Hospital A wants to run a notebook that computes joint statistics on shared patient data. They can now upload and run that notebook themselves — but only after Hospital B explicitly approves the code. This balances flexibility with governance: data never moves without consent, code never runs without review, and output always stays with the runner. For customers, this means faster iteration, more autonomy, and full auditability — all within a trusted collaboration environment.

Code governance is critical in data collaboration. These approval features ensure no code runs without consent. You gain an audit trail of who approved what (no more surprise queries), and you reduce the risk of malicious or erroneous code.

Walkthrough: Putting It All Together

Let’s walk through a high-level example using all three features:

1. Create the Clean Room (on GCP): Data Platform Lead (let’s call him Alex) sets up a new clean room on Google Cloud. She invites two collaborators by sharing their identifiers (AWS and Azure partners)
2. Add Data: Each party adds tables to the clean room. For example, the AWS, GCP, and Azure datasets are all uploaded in the clean room. There are now three collaborators in this room—Alex and two invitees. All three can see the shared metadata, but none can see others’ raw data.
3. Add Notebooks: Alex uploads analysis notebooks (e.g., Python scripts for ML, SQL for joins). All parties must explicitly approve the notebook before Alex can run it.
4. Run the Analysis: Alex now kicks off the notebook run. Because it’s approved, the serverless clean-room cluster executes the code. The run produces output tables (temporary read-only results) in Unity Catalog. For example, a joined customer segment table is generated so the notebook executor can extract aggregate insights.
5. Review and Iterate: Alex views the aggregated outputs. If new analysis is needed, they repeat, adding or updating notebooks under the same governance.

This demo shows how cross-cloud partners (on GCP, AWS, and Azure) can collaborate in a single shared clean room. It highlights multi-party collaboration, seamless data sharing via Delta Sharing, and the ability for partners to upload and run their own notebooks, with explicit approvals. The result: secure, auditable joint analysis across clouds and organizations, with no raw-data exposure.

What’s Next?

Databricks Clean Rooms continue evolving, but the core value remains: making data collaboration possible without compromising privacy, performance, or platform flexibility. We invite you to explore these new capabilities and share feedback. Ready to dive deeper? Check out the following sessions to learn more about Clean Rooms at the Data and AI Summit, June 9-12, 2025, in San Francisco

• Using Clean Rooms for Privacy-Centric Data Collaboration
• What’s new with Collaboration: Delta Sharing, Clean Room, Marketplace and the Ecosystem
• Intuit's Privacy-Safe Lending Marketplace: Leveraging Databricks Clean Rooms
• Unlocking TV Viewership Insights: Privacy-Compliant Data Sharing with Clean Rooms