What Is a Neural Network?
A neural network is a computing system inspired by biological neural networks that learns to perform tasks by processing examples rather than following explicit rules. It consists of layers of interconnected nodes (neurons) that receive inputs, apply mathematical transformations, and pass outputs to the next layer. During training, the network adjusts the strength of connections between neurons to minimise errors in its predictions. Deep neural networks — those with many layers — power most modern AI applications including image recognition, language understanding, speech synthesis, and autonomous systems. The term deep learning refers to machine learning using deep neural networks.
Neural Network — a class of machine learning model loosely inspired by biological neural systems, composed of interconnected processing units (neurons) organised in layers that transform input data through learned weights.
Neural networks are the architectural foundation of deep learning. Modern frontier AI — large language models, image generators, and multimodal systems — is built on transformer-architecture neural networks with hundreds of billions of parameters. From a governance perspective, the implication is that neural networks are typically opaque (the "black box" property), making explainability and interpretability standing concerns and an active research area.
Source: ISO/IEC 22989; NIST AI 100-1 (AI RMF)
Why it matters for governance
Neural networks create the 'black box' governance challenge: as networks grow deeper and more complex, understanding why they produce specific outputs becomes increasingly difficult. This lack of explainability conflicts with regulatory requirements for transparency (GDPR Article 22, EU AI Act, UK DUAA). Governance must address the trade-off between model performance and explainability, implement post-hoc explanation techniques where needed, and ensure human oversight mechanisms can effectively challenge neural network outputs in high-risk decision contexts.