A factory adds an AI assistant that reads machine data and can send new setpoints to the SCADA system, with a human approving every change. Sounds safe? Nathan Pembe and Maxime Tisseghem (NVISO) show why it is not. Together with the audience, they threat model this setup live, step by step, and turn it into a real attack plan. You leave with a repeatable method to threat model any system where AI meets physical processes. Free entrance, followed by networking with drinks.