Budibase bootstrap bypass is a tenant-scoping flaw in self-hosted, single-tenant deployments. Budibase’s GitHub advisory, published on November 10, 2026, describes unauthenticated creation of new tenant databases and administrator records after initial setup. Operators should identify reachable instances, confirm the running build, and review unexpected CouchDB growth before changing deployment mode or restoring tenant data.
The CVE Program published CVE-2026-122247 on November 20 and rated it 8.2 High under CVSS 3.1. The advisory’s demonstration did not establish login to the existing default tenant while multi-tenancy remained disabled. That limit matters when deciding what to investigate and what to report as confirmed.
What Is the Budibase Bootstrap Bypass?
Bootstrap is the first-run process that creates an administrator for a new installation. After provisioning, a guard should prevent that process from running again. The flaw occurs because the guard checks whether a user exists inside the tenant context selected for the request, rather than enforcing the single legitimate tenant.
On a single-tenant instance, a previously unused tenant context appears empty to that guard. The public setup function can then create a separate CouchDB database and global administrator document without a session. This is a persistent unauthorized write and a potential resource-exhaustion path, according to the Budibase advisory.
- Reachable setup functionAn unauthenticated request reaches the first-run check.
- Tenant contextThe check uses a newly selected namespace.
- Empty user checkA fresh namespace appears uninitialized.
- Persistent writeA database and administrator record are created.
This diagram summarizes the advisory’s trust-boundary failure without reproducing a request. It does not show access to applications or data in the established default tenant. The distinction guides both incident assessment and the order of defensive checks.
Which Self-Hosted Instances Need Review?
The documented precondition is a self-hosted Budibase deployment configured for single tenancy with its worker API reachable through the front end. An internal-only instance still needs a version review, although network reachability changes who could trigger the flaw. Record the actual running build and configuration for each instance, including replicas and restored environments.
The published version guidance conflicts. Budibase’s GitHub advisory marks the package below 4.0.0 affected and 4.0.0 patched. The CVE Program record instead lists versions before 3.50.0 as affected and says both 3.50.0 and 4.0.0 contain the fix.
| Source | Affected range stated | Fix stated |
|---|---|---|
| Budibase GitHub advisory | Below 4.0.0 | 4.0.0 |
| CVE Program record | Before 3.50.0 | 3.50.0 and 4.0.0 |
Do not silently treat the two ranges as identical. For a 3.50.x installation, ask Budibase to confirm whether the exact deployed build contains the fix. Keep the instance in the review queue until its version, deployment mode, and network exposure are documented.
Use the Vulnerability Intelligence platform as the product context for prioritizing this exposure across an estate. The source conflict still requires a build-specific answer from Budibase. A product inventory cannot resolve that discrepancy by itself.
What Did the Advisory Demonstrate, and What Remains Conditional?
The researcher’s test showed creation of a new tenant database and a global administrator document. Repeating that operation with different tenant contexts could consume CouchDB resources, including storage and database metadata. The record therefore supports a denial-of-service risk and persistent unauthorized changes.
The same test could not use the new administrator to log into the existing default tenant while multi-tenancy was disabled. Budibase’s advisory describes later multi-tenant operation, restore, or migration as circumstances that could make the rogue record relevant. Treat those as conditional risks, rather than proof that an existing tenant was taken over.
Brandefense Blackhole counted 29 dark-chat posts matching the broad term Budibase from August 23 through November 21, 2026. The measured unit is matching posts, which may discuss unrelated subjects. This count does not identify victims, establish discussion of this CVE, or verify exploitation.
The CVE’s High rating reflects its potential impact, not a count of affected installations. A reachable, vulnerable instance deserves prompt review because an unauthenticated actor could leave persistent records and create database load. Investigation should still separate observed local changes from the advisory’s possible outcomes.
How Should Teams Check Their Budibase Estate?
Start with an inventory of self-hosted instances and their actual running versions. Confirm whether each is configured for single tenancy and whether untrusted networks can reach its front end. Include instances used for testing, recovery, and migration because those environments may retain data or be exposed differently.
Next, compare CouchDB database counts and growth with authorized provisioning history. Review tenant administrator records for entries your team did not create, using a controlled administrative process. Preserve relevant access and datastore records before cleanup so responders can distinguish expected setup from unexplained writes.
Then restrict untrusted reachability while you verify the applicable fix with Budibase. Apply a confirmed fixed release through your patch management process, restart the service, and check the running build. A changed deployment manifest alone does not prove that every instance is running the intended version.
Finally, review unexpected tenant records before enabling multi-tenancy or restoring tenant data. If records are unexplained, investigate their creation time, related access activity, and any datastore pressure. Escalate on local evidence; the advisory does not supply an incident count for your environment.
- InventoryRecord running build, single-tenant configuration, and network reachability.
- Validate fixConfirm the exact 3.50.x or 4.x build with Budibase.
- Inspect datastoreCompare tenant databases and administrator records with authorized changes.
- Contain and updateRestrict untrusted access, update, restart, and verify the running build.
- Preserve evidenceKeep relevant access and CouchDB records before cleanup.
How Is the Separate Builder XSS Different?
Budibase also published GHSA-r2q3-7m4h-63qg on November 10; it became CVE-2026-122244 on November 20. That issue is stored cross-site scripting in a Builder preview, not a bootstrap guard bypass. Its described path requires malicious stored content and a later builder or administrator interaction.
The Builder issue’s GitHub advisory also labels versions below 4.0.0 affected and 4.0.0 patched. Its CVE Program record states versions before 3.50.0 and names 3.50.0 and 4.0.0 as fixed. Track its separate exposure and fix confirmation without combining its impact with the bootstrap flaw.
For the bootstrap case, the first decision remains concrete: identify reachable single-tenant installations and verify their running versions. Then compare tenant and CouchDB changes with authorized activity. The CVE record guide explains why a CVE description and a linked advisory may need separate review.
Frequently Asked Questions
Which evidence should a self-hosted team preserve before cleanup?
Preserve the running build, deployment configuration, access records, CouchDB database history, and tenant administrator records. Compare changes with authorized provisioning.
Should teams change to multi-tenancy before reviewing unexpected tenant records?
Review unexplained records first. The advisory describes later multi-tenant operation or migration as a conditional route by which rogue records might matter.
Is the Builder XSS the same vulnerability?
No. CVE-2026-122244 concerns stored content displayed in Builder preview. CVE-2026-122247 concerns the single-tenant bootstrap guard and persistent database writes.
