AI‑generated 'vibe coding' raises security concerns

Vibe coding lets users describe the desired function of an app in plain language and have an AI chatbot generate the code, fix errors and produce a working product. The term was coined by AI researcher Andrej Karpathy, and tools such as OpenAI's ChatGPT, Anthropic's Claude and Google Gemini are commonly used.

A recent Stack Overflow survey shows 84 % of developers plan to use AI coding tools in 2025, while about 63 % of vibe‑coding users have no programming background. Researchers, however, have uncovered thousands of security flaws and exposed secrets in publicly available AI‑generated applications, warning that software that appears to work may not be trustworthy. Proper testing, code review and maintenance remain essential before releasing AI‑generated software.

The trend promises to lower barriers to software creation for millions, but the security and reliability risks could affect a wide range of users and developers worldwide.