Snowflake SQLAlchemy versions before 1.11.0 expose multiple injection flaws: the library does not correctly sanitize user‑supplied column identifiers used in MERGE statements, it improperly renders bound parameters when constructing Snowflake‑specific CREATE TABLE queries, and it forwards connection configuration parameters in a way that permits reading local files. These weaknesses can allow an attacker to inject arbitrary SQL, alter data, or exfiltrate information that the application’s database role can access, and to read sensitive files from the application host and send them to an external endpoint.

The vulnerability affects installations of the Snowflake SQLAlchemy Python package older than 1.11.0. Any application that integrates this library—particularly dynamic upsert endpoints, table‑creation APIs, or connection string handling routines that accept user input—is potentially exposed.

The CVSS score is 8.3, indicating high severity, while the EPSS score is below 1%, suggesting a very low probability of exploitation in the wild. The vulnerability is not listed in CISA KEV. Attackers would need to supply crafted request fields or connection parameters to an application that uses the vulnerable package, typically via standard web or API calls, and would not require privileges beyond those granted to the application’s database role.