Massive LiteLLM Credential Archive Remains Exposed
A 153GB data leak from the 2026 LiteLLM breach continues to expose sensitive credentials for thousands of global organizations.
Updated on Oct. 21, 2026 in Cybersecurity
Live Poll
Do you trust that the software tools used by your workplace are adequately secured against hacking?
A 153GB archive containing 433,909 files from 2,488 organizations remains accessible following a March 2026 breach of LiteLLM. The breach, orchestrated by the group TeamPCP, compromised major entities including Nvidia, AWS, Cisco, Samsung, and Siemens.
Why it matters
The breach targeted LiteLLM due to its central role as a gateway for AI model API calls, allowing attackers to harvest critical infrastructure secrets. Despite arrests, the continued availability of these credentials poses an ongoing risk to enterprise security.
The archive includes credentials from 2,488 organizations, while malicious code injected into Python .pth files enabled the automated harvesting of SSH keys, database passwords, and Kubernetes secrets.
The players
TeamPCP
This cyberattack group was responsible for exploiting pipeline access at multiple security and development platforms.
Ruben Thomson
He is one of two men from Perth charged with 14 offenses related to the TeamPCP cyberattacks.
Louis Gaebler
He is an Australian man charged alongside Thomson for his involvement in the credential theft operations.
FBI
The Federal Bureau of Investigation is the U.S. agency that released security guidance regarding the rotation of compromised CI/CD tokens.
The details
Attackers utilized a poisoned security scanner to hijack PyPI publishing keys and push malicious LiteLLM versions 1.82.7 and 1.82.8. By leveraging these trojanized releases, the group exfiltrated sensitive environment variables from corporate development pipelines.
Timeline
March 19, 2026: Attackers pushed malicious commits into Trivy tags.
March 23, 2026: Attackers compromised Checkmarx KICS.
March 24, 2026: Attackers compromised LiteLLM and pushed trojanized releases.
July 2, 2026: The FBI issued an advisory urging the rotation of CI/CD secrets.
August 2026: Researchers confirmed credentials in the archive remained valid.
The Tech Race
The breach highlights the systemic risks inherent in widely adopted open-source AI integration tools like LiteLLM, which sees 95 million monthly downloads. This event marks a shift where supply chain attacks are increasingly targeting the connective tissue of modern AI development.
The exposure of these credentials necessitates immediate credential rotation for any organization that utilized LiteLLM during the compromise period. Developers should audit their CI/CD environments for any suspicious commits or unauthorized access to secret management stores.
The takeaway
The continued validity of stolen data highlights the critical need for immediate credential rotation following any third-party software compromise. Organizations must treat all integrated development tools as potential attack vectors for their most sensitive environment variables.
Further reading
For more information on securing development pipelines, visit the Cybersecurity section.
Live Poll
Do you trust that the software tools used by your workplace are adequately secured against hacking?
