Generative AI is rapidly changing the way software is developed. Rather than code directly in some programming language, developers are increasingly describing the functionality they want in natural language and using large language models to generate the concrete code. The goal of this report is to quantify the security properties of AI-generated code across a range of languages and models.