Through a compromised repository of the open-source tool LiteLLM, cybercriminals were able to capture the access data for build pipelines from over 2500 companies. The attack lasted only 40 minutes, during which the attackers extracted several terabytes of data.
Through a compromised repository of the widely used open-source tool LiteLLM, attackers were able to capture several terabytes of data and login credentials. Security company CloudSEK concludes, after analyzing the data, that the attack conducted in March of this year affected more than 2,000 companies, including major players like Microsoft, Nvidia, AWS, and Deutsche Bahn.
Security researchers have released new findings about a supply chain attack on the open-source tool LiteLLM. According to IT security companies CloudSEK and Hudson Rock, the incident affected more than 2,500 organizations. The data reportedly includes access credentials from companies such as Amazon Web Services, Cisco, Samsung, Salesforce, Siemens, Volkswagen, and Deutsche Bahn.
LiteLLM is an open-source library that can be accessed via a proxy server or Python SDK. It provides a unified interface to call over 100 large language models, such as those from OpenAI or Anthropic.
The starting point of the attack was manipulated versions of LiteLLM, which were distributed via the official Python Package Index (PyPI) in March. According to researchers, the compromised packages were available for only about 40 minutes. However, during this time, they were downloaded and executed by automated development and build systems in numerous companies.
CloudSEK and Hudson Rock estimate that data from around 434,000 CI/CD pipelines is affected. CI/CD systems automate tasks such as building, testing, and deploying software, often requiring extensive access rights to source code repositories, cloud services, and other internal systems. The available data indicates that corresponding access credentials may have been exposed. However, it does not confirm in every single case that all included credentials were actually used by attackers.
According to Hudson Rock, the findings were obtained in part through the analysis of a dataset totaling 195 terabytes. Among other things, the researchers discovered cloud keys, repository tokens, SSH keys, Kubernetes secrets, credentials for publishing software packages, environment variables, and keys for AI services. Some of this information could be linked to specific companies, while other datasets could not be attributed due to a lack of distinct identifiers.
Malware Extracted Data from Build Systems
The compromised software packages contained malicious code capable of reading information from the memory of affected systems and transmitting it to attacker-controlled infrastructure. According to CloudSEK, the malware, referred to as "Sandclock" by Google, specifically targeted API keys for language models and configurations of AI gateways. The data was reportedly encrypted before transmission.
According to the researchers, the LiteLLM incident is part of a larger attack campaign targeting software supply chains. Previously, the widely used vulnerability scanner Trivy had also been compromised. The software KICS and the Python SDK of the telecommunications provider Telnyx are also said to have been affected. The group TeamPCP claimed responsibility for the campaign, and security researchers consider their involvement plausible based on the current findings.
The particular risk of such attacks lies in the automated processing of software dependencies. If a package in a central source is manipulated, build systems can retrieve and execute the altered version without manual intervention. As a result, even a relatively short time window can be sufficient for malicious code to spread across numerous independent development environments.
It remains unclear how many of the exposed credentials can still be used. According to the researchers, simply removing the compromised package versions is not sufficient. Access keys, tokens, and passwords may still be valid unless explicitly revoked or replaced. According to CloudSEK, the FBI warned in July that actors linked to TeamPCP could exploit stolen credentials even long after the original attack.
Researchers Recommend a Comprehensive Exchange of Access Data
CloudSEK and Hudson Rock recommend that organizations using LiteLLM in their development environments specifically review versions 1.82.7 and 1.82.8. Potentially affected companies should assume, in their assessment, that secrets accessible to the respective LiteLLM environment may have been compromised. These include, among other things, cloud keys, Kubernetes service account tokens, as well as personal access tokens for GitHub and GitLab.
Date: 08.12.2025
Naturally, we always handle your personal data responsibly. Any personal data we receive from you is processed in accordance with applicable data protection legislation. For detailed information please see our privacy policy.
Consent to the use of data for promotional purposes
I hereby consent to Vogel Communications Group GmbH & Co. KG, Max-Planck-Str. 7-9, 97082 Würzburg including any affiliated companies according to §§ 15 et seq. AktG (hereafter: Vogel Communications Group) using my e-mail address to send editorial newsletters. A list of all affiliated companies can be found here
Newsletter content may include all products and services of any companies mentioned above, including for example specialist journals and books, events and fairs as well as event-related products and services, print and digital media offers and services such as additional (editorial) newsletters, raffles, lead campaigns, market research both online and offline, specialist webportals and e-learning offers. In case my personal telephone number has also been collected, it may be used for offers of aforementioned products, for services of the companies mentioned above, and market research purposes.
Additionally, my consent also includes the processing of my email address and telephone number for data matching for marketing purposes with select advertising partners such as LinkedIn, Google, and Meta. For this, Vogel Communications Group may transmit said data in hashed form to the advertising partners who then use said data to determine whether I am also a member of the mentioned advertising partner portals. Vogel Communications Group uses this feature for the purposes of re-targeting (up-selling, cross-selling, and customer loyalty), generating so-called look-alike audiences for acquisition of new customers, and as basis for exclusion for on-going advertising campaigns. Further information can be found in section “data matching for marketing purposes”.
In case I access protected data on Internet portals of Vogel Communications Group including any affiliated companies according to §§ 15 et seq. AktG, I need to provide further data in order to register for the access to such content. In return for this free access to editorial content, my data may be used in accordance with this consent for the purposes stated here. This does not apply to data matching for marketing purposes.
Right of revocation
I understand that I can revoke my consent at will. My revocation does not change the lawfulness of data processing that was conducted based on my consent leading up to my revocation. One option to declare my revocation is to use the contact form found at https://contact.vogel.de. In case I no longer wish to receive certain newsletters, I have subscribed to, I can also click on the unsubscribe link included at the end of a newsletter. Further information regarding my right of revocation and the implementation of it as well as the consequences of my revocation can be found in the data protection declaration, section editorial newsletter.
Additionally, the researchers advise examining log data for suspicious access and increasing control over outgoing network traffic from build systems. The incident also highlights the impact that compromised components can have in modern software supply chains: highly automated development processes can allow a single manipulated dependency to reach a large number of companies within a short time.
Organizations identified with high confidence as affected by CloudSEK and Hudson Rock