FlagThis

Abstract

As AI agents and LLMs accelerate the software development lifecycle (SDLC), they simultaneously expand the attack surface by introducing insecure code patterns, increasing the risk of secret leakage, and exacerbating dependency vulnerabilities. This session explores the dual nature of AI in development, shifting the focus from manual implementation to prompt-based engineering. It provides actionable strategies for "securing AI with AI," specifically through the use of custom security agents, specialized developer instructions, and automated guardrails to ensure that rapid, AI-driven development remains production-ready and secure.