Building Secure and Private AI Products: A multidisciplinary masterclass
This repository holds the notebooks, exercises and data for a one-day masterclass that teaches engineers and architects to build security and privacy into AI and machine learning products by design rather than adding it afterward.
The course material covers real-world threat modeling for AI systems, hands-on red teaming where a participant executes and evaluates attacks on models, meta prompt engineering and guardrails for writing privacy-aware prompts and catching insecure prompts or outputs, and data-flow analysis and risk assessment for choosing appropriate identification, sanitization and pseudonymization controls across a data pipeline. A further section covers building evaluation datasets and integrating security and privacy testing into a deployment workflow.
It accompanies a conference masterclass and is structured as exercises, supporting data and templates rather than only slides, intended to move a participant from theoretical concepts to practical, hands-on skills within one intensive session.
It is distributed as Jupyter notebooks with pinned dependency files, tested against Python 3.12, and released under the CC-BY-SA-4.0 license.