The 15th iteration of the “Building Security in Maturity Model” (BSIMM) report was just released. BSIMM is a data-driven model that helps organizations evaluate and improve their software security. One of the prominent trends in BSIMM15 is the increased use of AI/ML and large language models (LLMs) in software development.

The BSIMM15 study measured how best-in-class software security firms are working to secure AI. Since AI is still relatively new to software development, there is a lot of uncertainty regarding how to secure software in the AI era. In this talk, we cover the problems and best practices associated with securing AI.