Executive Summary
The Hacker News reported that MCP servers can expose enterprise secrets through plaintext config, excessive permissions and prompt injection, and three high-severity Hugging Face Diffusers flaws let crafted model repositories run code when loaded.
What Happened
The Hacker News reported that MCP servers can expose enterprise secrets through plaintext configuration, excessive permissions, and prompt injection. It also reported three high-severity flaws in Hugging Face’s Diffusers library that could allow crafted model repositories to execute arbitrary code on systems loading them, creating an AI software-supply-chain risk.
Affected Products
- Hugging Face Diffusers
Vulnerability / Attack Technique
Crafted model repositories; MCP secret exposure
Technical Mechanism
As reported, loading certain model repositories with Diffusers could execute code.
AI Security Category
LLM03:2025 Supply Chain
Part of the MCP Security topic.
Part of the AI Supply Chain and AI SBOM topic.
Relevant Frameworks
RealGround Analysis
Loading a model is loading code. RealGround analysis: update Diffusers and only load models from trusted sources.
Recommended Controls
- Update Diffusers
- Load models only from trusted repositories
- Move MCP secrets to a secrets manager
Related RealGround Service
AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.
Sources
The Hacker News: https://thehackernews.com/search/label/Supply%20Chain%20Security?m=1&hl=ru
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.