Model Context Protocol

**Executive summary:** This report examines security risks introduced by the Model Context Protocol (MCP), explaining how malicious or misconfigured MCP servers, prompt-injection, and over-privileged tool exposure can enable credential theft, data exfiltration, unauthorized actions, and arbitrary code execution; it presents attack scenarios, detection/monitoring recommendations (least privilege, secure secret management, centralized logging, anomaly detection) and automated red-team testing via Giskard Hub.