Section 1.3 — Your Data Is Your Attack Surface

Most organisations focus their AI security concerns on the model itself — the algorithms, the outputs, the API. But in practice, the most exploitable part of an AI system is often the data that feeds it.

In this section we cover three connected risks: training data poisoning, data leakage from deployed models, and the supply chain risk introduced when you rely on third-party datasets, pre-trained models, or AI vendors.