video1.0<iframe src="https://www.loom.com/embed/7093b34c7e39479aaa0416803707afac" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/7093b34c7e39479aaa0416803707afac-61665c41e699f2f1.gif1260.967This Loom explains the governance, risk, and compliance implications of the nested “AI Family Tree” framework across four layers of technology. It starts with Layer 1 AI, noting AI definitions can be broad and that learning is optional, making legal compliance trigger if a system fits the definition, as regulators like the EU AI Act and NIST frameworks require jurisdiction. It then covers Layer 2 machine learning, emphasizing that training data can encode bias, illustrated by the UK A-levels grading algorithm issue, and Layer 3 deep learning as the “black box” problem needing explainability. Finally, it describes Layer 4 generative AI, including hallucinations, IP exposure, and data leakage, and warns about future “AI training on AI” compounding errors and bias.