Litellm

• Threat actor group “TeamPCP” orchestrated a sophisticated multi-ecosystem supply chain attack that compromised LiteLLM, a popular open-source AI proxy gateway. • The breach led to the silent exfiltration of sensitive developmental secrets from thousands of global enterprise CI/CD pipelines. • Cybersecurity firm Hudson Rock has independently obtained the raw exfiltrated data, providing an unfiltered view of the massive scale of the compromise.

• A supply chain attack by TeamPCP poisoned LiteLLM’s build pipeline via Trivy, potentially exposing 2,500 companies and 434,000 CI/CD pipelines. • The malicious payload targeted CI/CD runners to steal SSH keys, cloud credentials, Kubernetes tokens, and environment secrets from process memory. • The attack specifically sought LLM API keys and gateway settings, creating a direct path to sensitive data and connected AI systems.

• In March 2026, the AI infrastructure tool LiteLLM suffered a security compromise involving a malicious upload to the PyPI repository. • The breach transformed developer tooling into a conduit for high-value attacks, including cloud intrusion and the theft of sensitive credentials. • This incident highlights a growing trend where AI-specific software supply chains are becoming strategic targets for sophisticated cyberattacks.

• AI recruiting platform Mercor confirmed a cybersecurity incident from a supply chain attack on open-source project LiteLLM, with hacker group Lapsus$ claiming data access and theft. • The breach exposed Mercor's systems, highlighting vulnerabilities in AI supply chains used by startups for model integration. • This incident underscores rising risks for AI firms relying on third-party libraries, potentially impacting hiring trust and data security standards.