Catching a Vibe: Identifying Security Issues in Vibe-Coded Applications

No ratings

Presented at 0xcon 2025 by

Renico Koen

Generative AI coding assistants have changed the way developers write software. ""Vibe coding""—turning natural language prompts into working code in seconds—is now relatively common, bringing its own set of security challenges that require thorough analysis. This talk looks at the security of code from three major language model platforms. We assigned each system the same coding tasks and then ran the output through standard security tests. The goal of this talk is to provide security teams and developers with a realistic picture of the risks associated with AI-generated code. Understanding these patterns helps organisations develop more effective review processes. We'll share practical findings from our testing and discuss what this means for teams using AI assistance in their development process.