CVE-2026-45758 describes a supply chain attack where an attacker published a malicious version (0.10.1) of the guardrails-ai Python package to PyPI. This version contained embedded malicious code (CWE-506). The malicious package was live for about two hours before being quarantined. Although no telemetry or logs indicate exploitation or data exfiltration, the presence of malicious code in a widely used AI framework version poses a critical risk. The vulnerability has a CVSS 3.1 score of 9.6, reflecting network attack vector, low complexity, no privileges required, user interaction needed, scope changed, and high impact on confidentiality, integrity, and availability. Users are advised to avoid version 0.10.1 and use 0.10.0 or 0.10.2 instead.

The malicious package could potentially compromise confidentiality, integrity, and availability of affected systems due to embedded malicious code. However, no known exploits in the wild or evidence of data exfiltration have been observed. The critical CVSS score reflects the potential severity if the malicious code were executed. Users who installed the affected version risk exposure of sensitive credentials and unauthorized access if the malicious code was triggered.

Users should immediately upgrade to guardrails-ai version 0.10.2 or downgrade to 0.10.0, both confirmed unaffected. Those who installed version 0.10.1 should rotate all credentials accessible from their machines, including GitHub personal access tokens, cloud provider keys, package registry tokens, and API keys. Additionally, users should audit their GitHub accounts for unauthorized workflows or repositories. PyPI has quarantined the malicious package, preventing further installations. Patch status is not explicitly stated but upgrading or downgrading as above effectively mitigates the risk.