LiteLLM's Release Pipeline Hacked Again, Seeding 1.107.0 With Malware

Hackers seeded LiteLLM's 1.107.0 PyPI wheels and Docker images with credential-stealing malware in an October attack on the 92M-download project.

@sharedot

LiteLLM's Release Pipeline Hacked Again, Seeding 1.107.0 With Malware

By @sharedot · · 6 pages

  • Programming
  • Supply Chain Security
  • Python
  • Pypi
  • Open Source

Hackers seeded LiteLLM's 1.107.0 PyPI wheels and Docker images with credential-stealing malware in an October attack on the 92M-download project.

A second LiteLLM supply chain breach in seven months

BleepingComputer reports that hackers compromised the LiteLLM project's release pipeline on October 21, 2026, pushing version 1.107.0 to PyPI and Docker Hub and GHCR with compiled malware that steals credentials and loads remote-code execution tools.

LiteLLM is an open-source Python SDK and AI gateway with 61,800 GitHub stars and roughly 92 million monthly PyPI downloads, used by companies like NVIDIA, Netflix, AT&T, and Okta.

According to the maintainer's post-mortem, the compromised artifacts included all seven PyPI wheels for litellm==1.107.0 (live from 11:39:51 to 13:23 UTC), Docker Hub tags v1.107.0, latest, main-latest, and main-stable (from 11:46:56 to 14:20 UTC), and corresponding GHCR images that were made unavailable by 13:51 UTC.

Customers running the private enterprise version were not impacted.

The malware fires before your app even starts

The surprising part is where the payload hid: BleepingComputer reports the attacker never touched the LiteLLM application code, instead dropping three files — cook.pth, cook.py, and sitecustomize.py — into Python's library directory.

Socket researchers explained that Python processes .pth lines and imports sitecustomize at startup, so the payload runs before the application; Endor Labs noted that simply launching Python or pip to investigate an affected environment also triggered the malware.

The stolen data covered cloud and deployment credentials (AWS, Azure, Google Cloud, Kubernetes, Docker), AI tool authentication files and LiteLLM's stored provider credentials, developer secrets including Git, GitHub, npm, PyPI, CI/CD credentials, SSH, GPG and TLS keys, browser data, and environment variables.

Data was encrypted with an AES-256-CBC key wrapped by RSA-4096 and uploaded to 'litellm[.]dev/healthcheck'.

Detected in minutes, but the damage question remains

One mitigating detail: BleepingComputer reports that multiple application security companies detected the attack within six minutes of the malicious version being published and analyzed the payload.

Detection speed matters because a related March 2026 incident — described by Rescana — saw malicious versions 1.82.7 and 1.82.8 remain on PyPI for only about 40 minutes, yet the campaign, attributed to the threat actor TeamPCP, still harvested over 50 categories of secrets and exfiltrated them via HTTPS POST to models.litellm[.]cloud.

In that earlier breach, Rescana reports attackers compromised Aqua Security's Trivy scanner in the CI/CD pipeline and also used stolen GitHub CLI OAuth tokens from three engineers to approve a malicious release, spanning PyPI, npm, Docker Hub, GitHub Actions, and OpenVSX.

Recovery: rebuild, don't just uninstall

The malicious packages have been removed, but BleepingComputer reports that potentially impacted users should stop affected deployments and inspect environments without launching the Python interpreter, since running Python or pip executes the payload.

Infected systems should be isolated and rebuilt from clean backups, with all exposed secrets rotated immediately — removing the LiteLLM package directory alone may leave startup files or an active loader behind.

The recommended recovery version is LiteLLM 1.106.2, the latest clean release before the attack.

Rescana's analysis of the March incident urges auditing for malicious .pth files (litellm_init.pth, cook.pth) in site-packages, monitoring outbound traffic to attacker-controlled domains, pinning dependencies, using official Docker images, enforcing least privilege for developer credentials, and hardening CI/CD pipelines.

The stakes are high because AI gateways concentrate provider API keys and centralized secrets.

Sources

  1. bleepingcomputer.com › LiteLLM release pipeline hacked to distribute info-stealer malware
  2. rescana.com › LiteLLM Supply Chain Attack Analysis (March 2026)

More on Programming