The proxy that sat between your code and your LLM was compromised from the inside.

LiteLLM — the widely-used open-source proxy that developers use to manage connections to multiple LLM providers — suffered a supply chain attack that turned it into an exfiltration tool. The compromised package silently harvested SSH keys and AWS credentials from every system it was installed on.

Trend Micro's investigation revealed the attack had been carefully orchestrated: the malicious code was embedded in an update that looked legitimate, passed automated security checks, and deployed to thousands of production environments before anyone noticed.

The scale of exposure was immense. LiteLLM sits at the intersection of AI infrastructure and cloud credentials — the exact junction where an attacker gets maximum value. Every developer who updated their LiteLLM installation during the compromise window potentially exposed their most sensitive infrastructure secrets.

The AI supply chain is the new frontier of supply chain attacks. When your LLM tooling is compromised, the blast radius includes everything that tooling can access.