THE S IN MCP STANDS FOR SECURITY

Summary from the TLDR InfoSec newsletter

Model Context Protocol (MCP) is a new standard that defines how LLMs integrate with tools. It lacks authentication or other security integrations by default. Some risks associated with MCPs are injection risks due to poor coding practices, tool poisoning due to embedding malicious instructions inside the MCP tool's description, silently redefining tool definitions, and cross-server tool shadowing. Users and developers should continue to employ security fundamentals such as using trusted sources, employing input validation, and pinning MCP server + tool versions. Platform builders can enhance security by displaying full tool metadata, using integrity hashes for server updates, and enforcing session security.