CVE-2026-41490
8.3Dagster Labs · Dagster
Dagster I/O managers are vulnerable to SQL injection through unescaped dynamic partition key interpolation, allowing authenticated users with specific permissions to execute arbitrary database queries.
Executive summary
An SQL injection vulnerability in Dagster allows authenticated users with partition creation rights to execute arbitrary database queries, creating significant risk for connected data warehouses.
Vulnerability
This is an SQL injection flaw (CWE-89) where I/O managers for DuckDB, Snowflake, BigQuery, and DeltaLake construct SQL WHERE clauses by interpolating dynamic partition key values without proper escaping, requiring low privileges to exploit.
Business impact
A successful exploit could allow an authenticated attacker to execute arbitrary SQL statements against the target database backend, potentially leading to unauthorized data access, modification, or data destruction. Given the high CVSS score of 8.3, the flaw poses a severe threat to the confidentiality and integrity of enterprise data pipelines and associated data assets.
Remediation
Immediate Action: Update Dagster Core to version 1.13.1 and associated Dagster libraries to version 0.29.1 immediately.
Proactive Monitoring: Monitor database query logs for unusual SQL syntax or unauthorized access attempts originating from the Dagster I/O manager service accounts.
Compensating Controls: Restrict the Add Dynamic Partitions permission to trusted users only and implement database-level permission boundaries for the I/O manager service accounts.
Exploitation status
Public Exploit Available: Yes, a public proof-of-concept repository exists on GitHub (https://github.com/romain-deperne/CVE-2026-41490).
Analyst recommendation
Administrators must treat this vulnerability with high urgency due to the availability of a public proof-of-concept and the potential for severe database compromise. Immediate patching of Dagster Core and related libraries to the specified fixed versions is required to neutralize the threat.