LiteLLM Supply Chain Attack Exposes Government AI Workload Risk
Attackers hijacked LiteLLM's release pipeline and shipped version 1.107.0 through PyPI and Docker registries with a credential stealer and remote-access trojan. Federal contractors running LiteLLM in AI workloads, Python environments, containers, or CI runners should treat exposed credentials as compromised, rotate secrets from clean environments, and rebuild affected systems.