LiteLLM Compromised to Steal Cloud and AI Credentials
LiteLLM version 1.107.0 shipped malicious Python startup files that collect cloud credentials, private keys, browser data and auth files for AI tools (Codex, Claude, Gemini, Cursor); the payload activates on normal Python startup without importing LiteLLM. Socket reports the release workflow pointed at a lookalike repository (Berri-AI/litellm), and both the PyPI wheels and Docker images were affected — LiteLLM gets about 23 million weekly downloads. Anyone who installed 1.107.0 should isolate affected environments and rotate every credential that process could see. It matters because this is a supply-chain attack on one of the most-used LLM gateways: even starting Python to inspect the install can re-trigger the stealer.
SOURCES
- Hacker News (AI)2026-10-21 15:24