SlowMist alerts on LiteLLM supply chain compromise in version 1.107.0
SlowMist reported a supply chain attack on LiteLLM in which the malicious release [email protected] affected PyPI packages and Docker images, including floating tags such as latest and main-stable. According to the post, the injected malware runs automatically during Python startup, steals cloud credentials, AI API keys and SSH private keys, and enables remote code execution. SlowMist said affected releases have been removed or replaced and advised avoiding version 1.107.0, rebuilding compromised environments from clean images, rotating exposed credentials and pinning to 1.106.2.
Why “Unverified”
- Reported by SlowMist alone, with no second source or primary confirmation yet
1 independent source group across 1 post · evidence score 72/100. How labels work
What the sources claim
-
SlowMist says LiteLLM was compromised again in a supply chain attack via the malicious release [email protected].
“LiteLLM has been compromised again in a supply chain attack. The malicious release [email protected]”
SlowMist -
The malicious release affected PyPI packages and Docker images, including floating tags such as latest and main-stable.
“affects PyPI packages and Docker images, including floating tags such as latest and main-stable”
SlowMist -
The injected malware executes automatically during Python startup and steals cloud credentials, AI API keys and SSH private keys.
“The injected malware executes automatically during Python startup, stealing cloud credentials, AI API keys, SSH private keys”
SlowMist -
The malware also enables remote code execution.
“It also enables remote code execution.”
SlowMist -
Affected releases have been removed or replaced.
“Affected releases have been removed or replaced.”
SlowMist -
SlowMist recommends pinning to version 1.106.2.
“Pinning to 1.106.2 is recommended.”
SlowMist
Sources
- 🚨 SlowMist TI Alert 🚨 → misteye.io
More in Hacks & Security
Summary and claims extracted by AI with prompt analyze-story@5 on 2026-10-23 00:00 UTC. Labels are computed from the sources, not by the model.