CVE-2026-45758

Affected versions of the guardrails-ai package are vulnerable to Embedded Malicious Code due to a supply-chain compromise in which an unauthorised party published a tampered release to PyPI. Version 0.10.1 of guardrails-ai was uploaded to PyPI as a malicious build that diverges from the legitimate upstream guardrails-ai source tree and contains code executed at install or import time, distinct from any sanctioned release of the project. Any environment that installed guardrails-ai 0.10.1 from PyPI during the window between publication and PyPI quarantine should be treated as compromised, as the embedded payload can perform arbitrary actions on the host with the privileges of the installing or importing process, including credential theft and onward lateral movement.