LiteLLM 1.107.0 Was Poisoned for a Few Hours: How to Tell If Your AI Stack Was Hit

Many Indian teams use LiteLLM as a single gateway to call different AI model providers. On 21 October 2026, a malicious LiteLLM 1.107.0 was published to PyPI and Docker registries. It contained malware designed to steal credentials. LiteLLM and Datadog Security Labs both reported the incident. The bad packages came from a lookalike repository, while the official GitHub source tag stayed clean.
The risky window (21 October, UTC)
- PyPI litellm==1.107.0, 11:39 to 13:23
- Docker Hub tags v1.107.0, latest, main-latest, main-stable, 11:46 to 14:20
- GitHub Container Registry images, from 11:46 until removed by 13:51
If anyone installed or pulled LiteLLM without a pinned version during those hours, treat that machine as a suspect.
Why a simple upgrade is not enough
The malicious files can run again every time Python starts, so upgrading on the same host may leave them in place. The safer path:
- Rebuild affected images and environments from scratch.
- Pin version 1.106.2 or another release from before the incident.
- Rotate every key the host could reach: model provider keys, LiteLLM master and virtual keys, database passwords and cloud roles.
- Review billing and logs for unusual use since 21 October.
Read the full LiteLLM guide here: What Indian AI teams must check after the LiteLLM 1.107.0 compromise
Published by Affix Center, a Mumbai-based IT services company that supports cyber security and AI projects.
Comments
Post a Comment