Ricardo Goncalves - The Model Context Protocol: A Deep Dive into the New AI Security Attack Surface

This talk focuses on the Model Context Protocol (MCP) as a crucial layer for integrating AI assistants with various tools and services. The speaker examines how MCP changes traditional application security boundaries and addresses the emerging security risks that come from AI-driven context, delegated execution, and the use of third-party tools. Attendees will learn to identify MCP security vulnerabilities across three key areas: classic application and API threats, supply chain challenges, and unique risks related to the protocol's design and model-driven decision-making processes.