CVE-2026-81335

Baserow's Application Builder data-source dispatch and record-name API views (backend/src/baserow/contrib/builder/api/data_sources/views.py) are declared with a permission class that admits any caller, and DataSourceService.dispatch_data_sources runs check_multiple_permissions without raising and discards the returned denials, so the permission check result is never enforced (CWE-862). An unauthenticated attacker can call the dispatch or record-name endpoint with a data source ID; the identifiers are small integers and can be enumerated. The data source is then dispatched using the integration's own credentials, so the attacker receives the rows and fields that data source reads, a confidentiality-only impact scored 8.7 (High) in CVSS 4.0. Deployments running Application Builder data sources on Baserow versions before 2.3.1 are affected; version 2.3.1 passes raise_exception to the permission check and thereby closes the gap. No public proof-of-concept, no CISA KEV listing, and a 0.4% EPSS (36th percentile) indicate no known exploitation to date.

What to do: Upgrade Baserow to 2.3.1 or later, which passes raise_exception to the permission check so denial results are enforced. Where upgrading is not immediately possible, require authentication in front of the builder data-source API (e.g., at a reverse proxy or WAF) and inventory which Application Builder data sources could leak sensitive rows, since data source IDs are small integers an attacker can enumerate. Also review the credentials attached to integrations, because dispatch executes with the integration's own access.