Latenode

LiteLLM pulls poisoned 1.107.0 package and Docker tags

A compromised AI dependency can hit secrets and runtime environments before a library is ever imported. Teams need package and container-tag monitoring, pinned versions, and incident playbooks that include host checks and credential rotation.

LiteLLM pulls poisoned 1.107.0 package and Docker tags

LiteLLM said on October 21, 2026 that attackers compromised its release pipeline and pushed a malicious litellm==1.107.0 package to PyPI along with poisoned Docker tags.

According to the LiteLLM incident report, the bad PyPI release was available from 11:39 to 13:23 UTC. The affected container tags — v1.107.0, latest, main-latest, and main-stable — were exposed from 11:46 to 14:20 UTC before removal or repointing. LiteLLM also said customers using its private enterprise registry were not impacted and told users to pin 1.106.2 or another earlier explicit version.

The risk was broader than apps actively calling LiteLLM. The malicious release used cook.pth and sitecustomize.py, which let code run when the Python interpreter starts. That means an infected environment could execute the payload before an application ever imports the package. Datadog’s analysis describes the malware as a credential stealer with remote-code capability aimed at cloud, developer, browser, and AI-service secrets.

For teams that pulled the bad package or tags, this is an environment-response problem, not just a rollback: check hosts and containers for indicators of compromise and rotate reachable credentials.

When models and APIs shift keep the workflow running

Connect the models and apps from stories like this in one scenario. If a vendor changes routing, pricing, or availability, you update the workflow — you don't start over.

Start Free

Free forever plan. No credit card required.