What happened in the LiteLLM supply chain attack?
On October 21, 2026, a malicious litellm==1.107.0 was published to PyPI and to LiteLLM's container registries. In its incident report, LiteLLM says the artifacts were built from an attacker-controlled repository, and the official v1.107.0 source tag on GitHub doesn't contain the malicious files.
LiteLLM pulled the PyPI release, removed or repointed the container tags, revoked maintainer credentials and rotated its image-signing key on October 23. It says the initial entry point is not yet confirmed.
Which LiteLLM versions and images are affected?
- PyPI:
litellm==1.107.0(all seven wheels), live from 11:39 to 13:23 UTC. - Docker Hub:
v1.107.0,latest,main-latestandmain-stable, from 11:46 to 14:20 UTC. - GHCR: the same tags on
ghcr.io/berriai/litellm,litellm-databaseandlitellm-non_root, from 11:46 UTC until unavailable by 13:51 UTC. - LiteLLM says earlier versions are safe, and customers on the private enterprise release were not affected.
How to check if you're affected
- Run
pip show litellmin every environment that installs it unpinned, including CI images and notebooks. Version 1.107.0 means the host ran the release. - Search lockfiles and requirements files for
litellmwithout a pinned version, and container manifests for the tags above pulled on October 21. - Look for
cook.pth,cook.py,sitecustomize.pyorlitellm-1.107.0.dist-infoin site-packages, and for/tmp/remote_code_loader.lock. - Compare running image digests with the compromised digest LiteLLM lists. LiteLLM warns that finding no indicators does not prove a host is clean.
What to do now
- Pin
litellm==1.106.2or another explicit earlier version. - Rebuild affected images and virtual environments from scratch. LiteLLM advises against running
pip installfirst in an affected environment, because each Python start re-triggers the malware. - Rotate everything the host could reach: the instance or task IAM role, provider keys stored in the proxy database, master and virtual keys, the database password and secrets in the process environment.
- Check cloud audit logs for the role's credentials being used from outside your infrastructure around the install time.
Where Nullify fits
How Nullify Covers This Class of Attack
Nullify finds every repository and container image that pulls an affected package version and runs the upgrade as one campaign, scoped by CVE, package, repo or team, with owners, Slack escalation and SLA tracking. Your developers approve every merge.
Rotation of the credentials the release could reach stays with your team.
Frequently Asked Questions
Which LiteLLM version was compromised?
Version 1.107.0, published to PyPI and LiteLLM's Docker Hub and GHCR images on October 21, 2026. LiteLLM says earlier versions are safe and recommends pinning 1.106.2 or another explicit earlier release.
How do I know if I installed the malicious LiteLLM release?
Run pip show litellm and look for version 1.107.0, then check site-packages for cook.pth, cook.py or sitecustomize.py. For containers, compare digests with the one LiteLLM published. No indicators doesn't prove a host is clean.
Do I need to rotate credentials after the LiteLLM compromise?
Yes, if any host ran 1.107.0. LiteLLM says to treat it as compromised and rotate the IAM role, provider keys in the proxy database, master and virtual keys, the database password and environment secrets.
Were LiteLLM enterprise customers affected?
LiteLLM says customers running the private enterprise release were not affected. The compromise covered the public PyPI release and the public Docker Hub and GHCR tags during the October 21 windows.
Update Log
- Nov 16, 2026: page published with LiteLLM's affected tags, windows and rotation guidance.