Transom

AI supply chain risk

Also called: MCP supply chain, agent supply chain

Risk introduced through the external components an agent depends on — MCP servers, packages, models, registries, and their updates.

An agent stack pulls in many third-party parts: MCP servers from a directory, framework packages from a language registry, model weights or endpoints, prompt templates. Each is code or content you did not write, delivered through channels with varying integrity guarantees, and often auto-updating.

Why it matters

The MCP-specific twist is that a "server" is both a dependency and a live, remotely-controlled counterparty — so it carries classic supply-chain risk (typosquatting, ownership hijack, malicious update) plus runtime risks like rug pulls and tool poisoning. Pinning versions and definitions, verifying publishers, and scanning on every update are the baseline defenses.

Exposure map

Live counts of instances showing this pattern will appear here once the exposure map is collecting data.

Related terms

Added 2026-09-02. Last reviewed 2026-09-02. Definitions in this space are evolving; entries are dated so revisions stay legible.