LiteLLM supply-chain attack exposes terabytes of credentials

Original: Terabytes of credentials leaked in massive supply-chain attack

Why This Matters

This breach highlights critical security risks in AI toolchain supply chains, affecting hundreds of major enterprises globally.

A supply-chain attack on open source AI tool LiteLLM exposed credentials from 2,500+ organizations including Microsoft, Amazon, Cisco, Samsung, and Salesforce. Security firms CloudSEK and Hudson Rock revealed the breach affected 434,000 CI/CD pipelines during a 40-minute window in March 2026.

Security firms CloudSEK and Hudson Rock disclosed on August 12, 2026 that a supply-chain attack on LiteLLM — an open source tool used to streamline AI-driven software development — resulted in the exposure of terabytes of credentials belonging to more than 2,500 organizations. The compromised data, contained in a 195TB file analyzed by Hudson Rock, includes cloud keys, SSH keys, Kubernetes secrets, repository tokens, package publishing credentials, environment variables, and AI provider keys. Major organizations affected include Microsoft, Amazon, Cisco, Samsung, and Salesforce.

The attack exploited compromised versions of LiteLLM distributed via the Python Package Index (PyPI). The malicious code accessed machine memory, scraped its contents, and exfiltrated the data through an attacker-controlled channel during a 40-minute active window in March. Approximately 434,000 CI/CD pipelines had credentials exposed as a result.

The LiteLLM compromise itself stemmed from a prior supply-chain attack targeting Trivy, a widely used vulnerability scanner. The campaign also infected KICS and the Telnyx Python SDK. The hacker group TeamPCP, described as a loosely organized but highly capable group largely composed of teenagers, claimed responsibility. Independent security researcher Kevin Beaumont confirmed the data's legitimacy, calling it 'a massive supply chain breach due to poor AI security' and noting that organizations rushing AI adoption with weak DevOps practices were the underlying vulnerability.

Source

arstechnica.com — Read original →