LiteLLM 1.107.0 supply-chain attack distributed credential stealer and backdoor
SafeDep found that all seven LiteLLM 1.107.0 wheels for Linux, macOS, and Windows installed three malicious files that could activate whenever a Python process started, even without importing LiteLLM. PyPI and container registries removed the affected release and images, while LiteLLM’s investigation said attackers used three stolen engineers’ GitHub credentials to alter the release workflow; the public source tag did not contain the malicious files. LiteLLM also raised the possibility that GitHub CLI OAuth tokens entered internal logs through Claude Code responses, but it had not confirmed the initial access route or whether Claude Code caused the token exposure.
The compromised package targeted AI-provider API keys and other cloud, browser, SSH, and development credentials, while also using the official release pipeline to distribute malicious code.
• Oct 21, 2026 •6 min read
On this page
The LiteLLM 1.107.0 release on PyPI carries a credential stealer and a remote code execution backdoor. All seven wheels (Linux, macOS, Windows) install them as three files in site-packages. Both run when any Python process starts on the host, even one that never imports litellm. The upstream GitHub source does not contain them. PyPI has since removed the release.
Versions 1.106.2, 1.107.0rc1, and 1.107.0rc2 are clean. The companion packages that 1.107.0 pins, litellm-proxy-extras==0.4.110 and litellm-enterprise==0.1.78, are also clean.
What the attack does
Installing litellm==1.107.0 places three files in site-packages. Python loads two of them each time it starts.
#LiteLLM#PyPI#Docker Hub#GitHub Container Registry#GitHub#Anthropic#Claude Code