Leaked LiteLLM master key used to steal GitHub tokens and publish malicious 1.107.0
An unrevoked Agent Platform key from May let an attacker publish LiteLLM 1.107.0 to PyPI on October 21; PyPI later quarantined it.
An unrevoked Agent Platform key from May let an attacker publish LiteLLM 1.107.0 to PyPI on October 21; PyPI later quarantined it.
October 10 and 12 workflows posted CI secrets to webhook.site; a key committed May 27 was never revoked.
Starting Python with the package installed can run PRIVCOLLECT without importing LiteLLM; removing the package does not recover credentials.
The poisoned build was online about three hours, downloaded over 49,000 times, and the latest clean release is 1.106.2.
Malicious LiteLLM 1.107.0 wheels and images were live for hours on October 21; roll back to 1.106.2 and rotate secrets.
The build stayed up for under two hours on PyPI; anyone who installed it is told to rotate credentials, not only upgrade.
Unpinned installs for a few hours on October 21 pulled a credential stealer and an hourly remote-code backdoor.
Malicious Python startup hooks can collect cloud and AI credentials even when applications never import LiteLLM.
PyPI and Docker removed the trojanized release within hours; the latest safe version named is v1.106.2.
All seven wheels and several Docker tags were tainted; PyPI removed the release, and Wiz calls 1.106.2 the latest safe build.
All seven wheels run the implants in every Python process; PyPI removed the release, and 1.106.2 plus two release candidates are clean.
PyPI quarantined the upload within about two hours; matching Docker tags included latest, main-latest and main-stable.
PyPI and official Docker tags carried the release for about two hours; the RAT dead-drop never delivered a live second stage.
Rapid7's Metasploit wrap-up adds 12 modules, including unauthenticated RCE for N-central, Langflow, LiteLLM, and dizqueTV.
CISA added seven actively exploited vulnerabilities to its KEV catalog, including SonicWall SMA1000 SSRF and command injection flaws.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 76 readers · 78 views · 34 reads · all time 167K views