LiteLLM 1.107.0 Compromised on PyPI: Credential Stealer Hidden in Library Downloaded 23 Million Times a Week

LiteLLM, the popular Python library used to route requests across AI models with about 23 million weekly downloads, was hit by a supply-chain attack on October 21 that shipped a credential stealer and a backdoor inside version 1.107.0. Security firms SafeDep, Socket, StepSecurity and Upwind all flagged the compromise. PyPI quarantined the malicious release within two hours, and the tainted Docker images were removed.
The attack was unusually dangerous. A file called cook.pth made Python run the malicious "cook" module at interpreter startup — so the stealer activated in every Python process on a host, even ones that never imported LiteLLM. It harvested cloud credentials, SSH keys, CI/CD secrets, AI developer-tool tokens, browser passwords and the LLM provider keys stored in a LiteLLM proxy database, encrypting and exfiltrating them to a lookalike domain. A second payload, sitecustomize.py, ran a background daemon that checked a GitHub dead drop every hour and downloaded and executed attacker code.
LiteLLM's security chief, Oliver Jensen, said the attackers used stolen GitHub CLI OAuth tokens from three engineers' accounts to trigger and approve an official-looking release build from a malicious fork named "Berri-AI/litellm" — one hyphen different from the real project. The incident is LiteLLM's second supply-chain compromise of 2026 after a March attack. Users who installed 1.107.0 are advised to revoke exposed credentials from a clean environment and pin version 1.106.2.
Source: read more
Comments
Post a Comment