What you will be able to do
- List the workspace and metastore prerequisites for using Clean Rooms
- Describe what a clean room creates and what each collaborator can and cannot see
- Explain the no-trust notebook approval model and how packaged clean rooms differ
- Apply the clean room safeguards, limits and audit sources when you design a collaboration
1.What Clean Rooms are and what they require
Sometimes two organizations want to run a joint analysis, such as overlapping customer lists, without either side handing its raw data to the other. Databricks Clean Rooms handles this case. It combines OpenSharing with serverless compute to create an environment where several parties can work on sensitive enterprise data without direct access to each other's data.
Because it is built on those two pieces, the prerequisites follow from them. You need a workspace enabled for serverless compute, a workspace enabled for Unity Catalog, and OpenSharing enabled on your Unity Catalog metastore. Storage location matters too, because clean rooms run on serverless compute. That compute can't reach an S3 bucket behind a storage firewall by default. You can keep the data in Databricks default storage or allow serverless access on the bucket. If the data is behind a firewall or private endpoint, SecureConnect lets clean room compute reach it without allowlisting its network.
Checkpoint 1 of 5· Check yourself
A workspace is enabled for Unity Catalog and serverless compute, but creating a clean room fails eligibility. Which missing prerequisite is the most likely cause?
Clean rooms move data with OpenSharing, so the metastore must have OpenSharing enabled. Serverless compute and Unity Catalog are the other two requirements.
“OpenSharing enabled for your Unity Catalog metastore.”Source: docs.databricks.com
2.What a clean room creates and who sees what
Creating a clean room creates three objects. There is a securable clean room object in your metastore and a matching one in your collaborator's metastore. In between sits the central clean room, an isolated, ephemeral environment that Databricks manages. It runs in a serverless compute plane in the cloud region chosen by the clean room's creator.
This is where Clean Rooms uses OpenSharing differently from a normal share. The tables, volumes, views and notebooks each collaborator adds are shared with the central clean room only, never directly with the other party. Each asset is registered there under its alias. Notebook code runs inside the central room. A notebook can write output tables, which give a collaborator read-only, temporary results in their own Unity Catalog metastore to work with in their own workspace.
Checkpoint 2 of 5· Check yourself
How does data move between two clean room collaborators?
OpenSharing carries each party's assets into the Databricks-managed central clean room. Participants never receive each other's data.
“Clean rooms use OpenSharing to share data securely to the clean room, but not between participants.”Source: docs.databricks.com
3.The no-trust model and notebook approval
The Clean Rooms model is described as "no-trust". Every collaborator has equal privileges, and that includes the creator, so neither party is an admin of the central clean room. Trust comes from code approval rather than from roles. You can upload a notebook yourself, but it runs only after the other collaborators approve it. In practice you run notebooks the other party uploaded, after reading the code. Running another party's notebook counts as approving it.
Notebooks are version-controlled, and only the latest version can run. That stops someone from changing code after it has been approved. The clean rooms system table shows which notebook version ran.
Packaged clean rooms are the exception. They use a provider-consumer model in which collaborators do not have equal privileges.
Checkpoint 3 of 5· Exam question
Two banks want to jointly analyze overlapping customer records without either side seeing the other's raw data, and both insist on equal control over which queries run and what results leave the environment. Select all statements that correctly describe how a Databricks Clean Room supports this requirement.(Select 3)
Correct answers: A, B, C — Both banks have equal privileges in the clean room, and running a new notebook or query requires approval from all collaborators.; The clean room runs computations in an isolated, ephemeral environment managed by Databricks rather than on either bank's own compute.; A collaborator can save read-only output tables from an approved query into their own Unity Catalog metastore instead of exporting raw rows.
- A. Clean rooms operate on a no-trust model where all collaborators hold equal privileges, and this mutual-approval requirement is exactly the mechanism that lets both banks control which analyses actually execute. Neither party can unilaterally run a new notebook against the shared data.
- B. Computations inside a clean room execute in an isolated, ephemeral environment managed by Databricks, keeping the workload separate from either collaborator's own production compute. This isolation is central to why neither bank's raw data or code is directly exposed to the other.
- C. Clean rooms let a collaborator persist the read-only output of an approved query as an output table in their own Unity Catalog metastore, rather than pulling raw rows out of the environment. This gives each bank a way to retain analysis results without breaching the no-raw-access model.
- D. There is no automatic conversion from a clean room into a Databricks-to-Open share based on collaborator count; a clean room is capped at ten collaborators and remains a clean room throughout its lifecycle. Databricks-to-Open is a separate, unrelated sharing mode.
- E. A clean room is locked down once created specifically to prevent new collaborators from joining later. Neither bank can add participants after that point, which protects the agreed-upon scope of the collaboration.
- F. Clean rooms exist precisely so that raw tables never have to be exposed to the other party; Lakehouse Federation is an unrelated feature for querying external databases and plays no role in how clean room data access is restricted.
Checkpoint 4 of 5· Check yourself
Your collaborator uploads a notebook. You have not clicked an explicit approve button, but you run the notebook from your workspace. What is the effect?
Running a notebook that another party added counts as approving it, so review the code before you run it.
“If you run a notebook added by another party, you implicitly approve the code.”Source: docs.databricks.com
Sources1
4.Safeguards, limits and auditing
Beyond notebook approval, the clean room's structure is fixed once it is created. Network and audit controls cover the rest. The table below lists the rules most likely to show up in a design question.
| Rule | Consequence |
|---|---|
| Locked after creation | No new collaborators can join an existing clean room |
| Ten-collaborator limit | Each clean room holds at most ten collaborators |
| Any collaborator deletes it | The central clean room is void and no tasks can run for anyone |
| No renaming | The name must be unique in every collaborator's metastore |
| Serverless egress control | Restricting the clean room blocks access to unauthorized storage |
| Cross-cloud or cross-region room | Only tables in default storage are accessible; volumes are not |
Audit records come from three places. The clean room events system table records actions by you and your collaborators. Your account audit log records clean room actions under the clean-room service. When a collaborator changes permissions on the clean room securable in their own metastore, that change is logged under the unityCatalog service. Comments added to a clean room securable stay in the workspace where they were written and are not passed to other collaborators.
They cannot be added. A clean room is locked once it is created, so new collaborators can't join. The parties would need a new clean room that includes all three from the start, up to the ten-collaborator limit.
Checkpoint 5 of 5· Check yourself
A security team needs to find every clean room action in the account audit log. Which service name should they filter on?
Clean room actions are logged under the clean-room service. unityCatalog captures only a collaborator's permission changes on the clean room securable in their own metastore.
“Clean room actions are also recorded in your account's audit log under the service clean-room.”Source: docs.databricks.com
Sources1
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.The collaborator who creates a clean room is its administrator and can run their own notebooks without anyone else's approval.Why is that wrong?
In a standard clean room all collaborators have equal privileges, the creator included. A notebook you upload yourself must be approved by the others before it runs.
Covered in The no-trust model and notebook approval
2.Every clean room follows the equal-privilege no-trust model.Why is that wrong?
Packaged clean rooms use a provider-consumer model instead, so collaborators' privileges are not equal.
Covered in The no-trust model and notebook approval
3.A volume in default storage can be used in any clean room, whatever region the central clean room is in.Why is that wrong?
Volumes in default storage are accessible only when the central clean room is in the same cloud and region. In a cross-cloud or cross-region room, only tables are accessible.
Covered in Safeguards, limits and auditing
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.https://docs.databricks.com/aws/en/clean-roomsOfficial docs
“a secure and privacy-protecting environment where multiple parties can work together on sensitive enterprise data without direct access to each other's data”
↩︎ What Clean Rooms are and what they require“enable SecureConnect so that clean room serverless compute can access it without allowlisting the clean room compute's network”
↩︎ What Clean Rooms are and what they require“are shared with the central clean room only, using OpenSharing”
↩︎ What a clean room creates and who sees what“Notebooks can also generate output tables that let your collaborator temporarily save read-only output to their Unity Catalog metastore”
↩︎ What a clean room creates and who sees what“While you can upload a notebook to yourself, it must be approved by the other collaborators before running.”
↩︎ The no-trust model and notebook approval“Only the most recent version of a notebook can be run.”
↩︎ The no-trust model and notebook approval“Neither party has admin privileges.”
↩︎ The no-trust model and notebook approval“After a clean room is created, it is locked down to prevent new collaborators from joining the clean room.”
↩︎ Safeguards, limits and auditing“If any collaborator deletes the clean room, the central clean room is void and no clean room tasks can be run by any user.”
↩︎ Safeguards, limits and auditing“Each clean room is limited to ten collaborators.”
↩︎ Safeguards, limits and auditing“you can use serverless egress control to manage outbound network connections”
↩︎ Safeguards, limits and auditing“All collaborators in a no-trust clean room have equal privileges, including the creator of the clean room.”
↩︎ Exam trap 1“They use a provider-consumer model in which collaborators do not have equal privileges.”
↩︎ Exam trap 2“In a cross-cloud or cross-region clean room, only tables in default storage are accessible; volumes are not.”
↩︎ Exam trap 3“OpenSharing enabled for your Unity Catalog metastore.”
↩︎ Checkpoint“they can see column names and column types, and they can run approved notebook code that operates over the data assets”
↩︎ Prediction“Clean rooms use OpenSharing to share data securely to the clean room, but not between participants.”
↩︎ Checkpoint“If you run a notebook added by another party, you implicitly approve the code.”
↩︎ Checkpoint“Clean room actions are also recorded in your account's audit log under the service clean-room.”
↩︎ Checkpoint - 2.https://docs.databricks.com/aws/en/opensharingOfficial docs
“Clean Rooms, a secure and privacy-protecting environment where multiple parties can work together on sensitive enterprise data”
↩︎ What Clean Rooms are and what they require“Enable OpenSharing on the Unity Catalog metastore that manages the data.”
↩︎ What Clean Rooms are and what they require“OpenSharing is also the basis for Databricks Marketplace, an open forum for exchanging data products, and Clean Rooms”
↩︎ What a clean room creates and who sees what