
Sign up to save your podcasts
Or


On March 24, 2026, the popular Python library LiteLLM—which provides a unified interface for various Large Language Models—was the target of a sophisticated supply chain attack. Malicious code was injected into versions 1.82.7 and 1.82.8 of the package on PyPI. The attack is attributed to a threat actor known as TeamPCP, who gained access to LiteLLM’s publishing pipeline by first compromising Trivy, an open-source security scanner used in LiteLLM's CI/CD process.
The compromise is particularly significant due to LiteLLM's wide adoption, with roughly 3.4 million daily downloads (nearly 97 million monthly).
The Attack Vector and Payload
The attackers used a poisoned Trivy GitHub Action to exfiltrate a PyPI publishing token, allowing them to upload malicious versions directly to PyPI while bypassing GitHub code reviews. The malware featured a three-stage payload designed for extensive data theft and persistence:
Discovery via "Fork Bomb"
The attack was discovered when researcher Callum McMahon noticed his machine became unresponsive due to an unintended "fork bomb". A bug in the malware's .pth file—which triggers on every Python interpreter startup—caused it to recursively spawn new subprocesses until system RAM was exhausted. Experts noted that if the attackers had better programming skills, the compromise might have gone undetected for weeks.
Immediate Remediation Steps
LiteLLM has since removed the poisoned versions from PyPI and rotated all maintainer credentials. Users are advised to take the following actions:
By Don RamónOn March 24, 2026, the popular Python library LiteLLM—which provides a unified interface for various Large Language Models—was the target of a sophisticated supply chain attack. Malicious code was injected into versions 1.82.7 and 1.82.8 of the package on PyPI. The attack is attributed to a threat actor known as TeamPCP, who gained access to LiteLLM’s publishing pipeline by first compromising Trivy, an open-source security scanner used in LiteLLM's CI/CD process.
The compromise is particularly significant due to LiteLLM's wide adoption, with roughly 3.4 million daily downloads (nearly 97 million monthly).
The Attack Vector and Payload
The attackers used a poisoned Trivy GitHub Action to exfiltrate a PyPI publishing token, allowing them to upload malicious versions directly to PyPI while bypassing GitHub code reviews. The malware featured a three-stage payload designed for extensive data theft and persistence:
Discovery via "Fork Bomb"
The attack was discovered when researcher Callum McMahon noticed his machine became unresponsive due to an unintended "fork bomb". A bug in the malware's .pth file—which triggers on every Python interpreter startup—caused it to recursively spawn new subprocesses until system RAM was exhausted. Experts noted that if the attackers had better programming skills, the compromise might have gone undetected for weeks.
Immediate Remediation Steps
LiteLLM has since removed the poisoned versions from PyPI and rotated all maintainer credentials. Users are advised to take the following actions: