AI agentic coding tools are redefining how developers build software — and how attackers exploit it. In this hands-on workshop, you’ll learn how to securely use tools like Cursor, Claude, and GitHub Copilot while identifying and mitigating AI-specific security risks. We’ll cover detection of tool poisoning, prompt injection, and toxic flow vulnerabilities in MCP servers, followed by AI-focused red teaming of LLM endpoints and AI-powered APIs to uncover jailbreaks, data leakage, and unsafe behaviors. Attendees will leave with practical skills, tooling, and security patterns for building safer AI-driven development workflows.

The presentation includes QR codes pointing to the technical resources needed to follow the demos. The presentation duration is ~45min