The Python package litellm, widely used in AI projects, suffered a critical security breach on PyPI. Versions 1.82.7 and 1.82.8 contain malicious code that exfiltrates cloud credentials, SSH keys, and Kubernetes secrets. The impact is massive: the package was downloaded 3,408,615 times yesterday alone, and it has accumulated 95 million downloads in the last month. If the environment was updated after March 24, 2026, keys should be renewed urgently, and it is recommended to audit all accesses from affected systems.
Your AI Stack Just Handed Over Your Root Keys: Inside the litellm PyPI Breach
Summary: The vulnerability in the compromised Python library litellm on PyPI exfiltrated cloud credentials, SSH keys, and Kubernetes secrets. Engineers should immediately remove the package and renew their keys.
Key facts
- litellm 1.82.7 and 1.82.8 versions contain malicious code
- Steals cloud credentials, SSH keys, Kubernetes secrets
- 3,408,615 downloads yesterday alone
- 95 million downloads in the last month
Why it matters
This vulnerability highlights the need for rigorous supply chain control to protect sensitive systems and data in AI projects. The rapid spread of the attack underscores the risk associated with automated updates.
@trendaisecurity
Embedded content for: Your AI Stack Just Handed Over Your Root Keys: Inside the litellm PyPI Breach