When identifier escaping is disabled via a non‑default configuration, the write_pandas() function in the Snowflake Connector for Python and its Universal Core variant incorporated DataFrame column names into generated SQL without escaping. This flaw can be leveraged to inject arbitrary SQL commands, allowing an attacker who can supply crafted column names to read, alter, or delete data that the application’s Snowflake role is permitted to access.

Snowflake’s Snowflake Connector for Python (versions prior to 4.8.1) and the Snowflake Connector for Python (Universal Core) (versions prior to 5.2.0) are affected. Users of older connector releases should verify their version and upgrade accordingly.

The CVSS score of 8.1 indicates a high‑severity vulnerability, and although an EPSS score is not provided, the lack of a KEV listing suggests no confirmed exploitation yet. The vulnerability requires that an attacker can supply column names to write_pandas(), which typically means control over user‑provided data or local code that constructs DataFrames. Thus, the attack vector is typically application‑level and may be mitigated by enforcing proper escaping or by limiting exposure of write_pandas().