Learn to build, break and secure AI from first principles with this hands-on workshop; from neural nets to attack classes & AI model exploration. No AI/ML experience is needed, just basic Python & command-line skills. All code provided. Ideal for builders, breakers and defenders. Learn how to attack and defend AI. Who is it for? – For hands on builders, breakers, cyber risk and security, audit. What are the requirements? – Python: basic familiarity; install modules, import them, call functions, write simple conditionals and loops (note: you do not need to be a programmer – this is not a coding workshop, you will be given code snippets to paste and adapt) – Basic command-line skills; editing text files, issuing and editing commands – no previous experience of AI or Machine Learning is required – bring your laptop with a recent python version installed (do a `pip install openai` to prove to yourself it works), make sure wifi works and you can add a hotspot What you will learn: – how neural networks work by creating a simple one, training it and observing its operation, “thought by thought” – the main attack classes, from black box (chat/training data), grey box (API access) to white box (access to the training and/or inference environment) – how to use an AI playground to rapidly explore and experiment with new AI models, including testing prompt injection and exploring safety layers – how to quickly fine-tune base models to accomplish specific tasks whilst defending against prompt injection – how to build and test security use cases with popular AIs, including from OpenAI and hosted Open Source AIs – use AI to generate project ideas for further practice to develop your skills