LiteLLM
Named by 71 kept items in this workspace.
29 items name it under classification.technologies, 26 under affected products and 16 under tools.
The items that name it 71
malicious PyPI release published in October; version not stated
October 21 release compromised; internet-facing internal LiteLLM deployment suspected as entry point
AI middleware library, in approximately 36% of cloud environments; subsequently compromised using credentials harvested from Trivy's CI/CD runners
95 million monthly downloads
targeted by PoeLLM malware
AI gateway exploit earned $40,000; additional exploit earned $4,250-$17,500
command injection exploited in the wild
Used with proxy farms in distillation campaigns, per Anthropic
AI infrastructure package with 95 million monthly downloads, backdoored by TeamPCP.
Open-source AI gateway providing a unified OpenAI-compatible API for over 100 LLM providers (OpenAI, Anthropic, AWS Bedrock, Azure, Google Vertex AI); present in approximately one-third of cloud environments per Wiz data
Local LLM proxy option for routing model SDK requests through an auditable local monitoring point
AI proxy routing model traffic; affected by MCP Gateway auth bypass (CVE-2026-59822) and MCP test endpoint command injection (CVE-2026-42271); also targeted for master key extraction from memory and default master key (sk-1234) abuse.
actively exploited pre-auth SQL injection exposing upstream API keys
AI gateway/proxy between applications and model providers; compromised via CVE-2026-42271 chained with CVE-2026-48710, exposing provider keys, master key, virtual keys, and database connection strings
Widely used AI infrastructure package, poisoned by Team PCP.
AI provider used for intelligent payload generation
Underlying model-provider abstraction dependency credited by the project
Compromised by Team PCP; described by CloudSEK as the largest supply chain attack targeting AI infrastructure to date, central to the AI supply chain angle of the campaign (157 organizations leaked OpenAI API keys).
PyPI package, two releases carried an identical credential stealer hidden in the published wheels
Proxy for pinning or dynamically selecting a model behind aurscan's openai backend
AI gateway present in over a third of monitored cloud environments; had four separate security events in six months: a supply-chain compromise, an SQL injection vulnerability exploited in the wild, a privilege escalation chain, and an authentication bypass.
Open-source AI gateway proxy giving a unified API over OpenAI, Anthropic, Azure OpenAI, Bedrock and other providers. The post's attack targets a victim LiteLLM gateway and also stands up a malicious LiteLLM gateway to intercept traffic.
PyPI package; maintainer account compromised by TeamPCP, malicious versions steal credentials from every Python process on host.
Open-source project that Strix builds on, used for LLM provider integration
OpenAI-compatible LLM server backend supported by OASIS
Compromised in related TeamPCP campaign
Supported LLM provider abstraction for Metis
LLM gateway supported by Shannon for reaching models via a custom base URL
Library providing multi-provider LLM support (OpenAI, Anthropic, etc.) and embedding calls (text-embedding-3-small) for the RAG tool optimizer
Gateway used to rewrite Anthropic identifiers to route requests to Z.AI's GLM 5.1 on an Anthropic-compatible endpoint
Gateway option for routing LLM traffic
PyPI/npm
proxy exposed publicly
Integration layer allowing DeepZero to connect to any LLM provider for LLM-assess pipeline stages
Open-source AI gateway and Python SDK for interfacing with LLM providers
- CVE-2026-42271: LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format
Proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format
In every workspace you can read 89
Security
AI
All technologies in this workspace.