LiteLLM suffers supply-chain attack via exposed master key

LiteLLM, an open-source tool used by AI developers, has confirmed a supply-chain attack resulting from an unrevoked master key. The key was accidentally exposed in a configuration file in May; although the file was removed, the key remained active.

An attacker used the key to gain control over LiteLLM’s GitHub repositories and distributed a malicious version of the tool (version 1.107.0) via the Python Package Index (PyPI). The compromised package contained credential-harvesting malware designed to run whenever the Python programming language starts on an infected machine.

Security firm Wiz reported that the malicious package was downloaded approximately 49,000 times. The attack follows a pattern known as ‘LLMjacking,’ where criminals use stolen cloud credentials to run AI services at the victim's expense. Evidence suggests attackers attempted to use stolen credentials to access Amazon AWS cloud services and AI models within hours of the breach.