Supply‑Chain Breach Hits Popular LiteLLM Python Library, Injects Credential‑Stealing Backdoor

A widely used open‑source library for large language model integration, LiteLLM, has been compromised in a supply‑chain attack that inserted both a credential‑stealing component and a persistent backdoor into version 1.107.0.

Security researchers from SafeDep, Socket, StepSecurity and Upwind reported that the malicious code was introduced during the package’s distribution process, affecting anyone who installed the compromised version from the Python Package Index (PyPI). The payload is designed to harvest authentication tokens and other sensitive data from applications that import the library, then relay the information to an external server under the attacker’s control.

LiteLLM, which simplifies the use of various LLM APIs for developers, has become a staple in many AI‑driven products and research projects. The breach therefore raises concerns for a broad range of downstream services, from chatbots to data‑analysis tools, that may now be exposing secrets without the users’ knowledge.

Supply‑chain attacks on open‑source software have risen sharply in recent years, exploiting the trust placed in package managers and automated build pipelines. In this case, the attackers appear to have tampered with the source distribution before it was uploaded to PyPI, a tactic that can evade typical static‑analysis checks if the malicious code is obfuscated or conditionally executed.

The maintainers of LiteLLM responded quickly, removing the tainted release and publishing a clean version. They also urged users to upgrade immediately, revoke any credentials that may have been captured, and audit logs for unauthorized access. Organizations that depend on the library are advised to verify the integrity of their environments and consider implementing stricter dependency‑verification practices, such as using hash‑based pinning or internal mirrors.

Industry observers note that the incident underscores the need for stronger security hygiene in the open‑source ecosystem, including better provenance tracking and more robust vetting of third‑party packages. As developers continue to integrate AI components at speed, the balance between rapid adoption and safeguarding the software supply chain remains a critical challenge.