Glossary
The European Union's comprehensive regulation classifying AI systems by risk level and imposing corresponding obligations on developers and deployers.
The EU AI Act and UK Approach
from AI & Law
The EU AI Act — which entered into force in August 2024 with phased implementation through 2027 — is the world's first comprehensive AI regulation. It classifies AI systems by risk level: unacceptable risk (banned outright, e.g., social scoring), high risk (subject to conformity assessments, human oversight, and transparency obligations), and lower-risk systems (lighter requirements). The UK has taken a deliberately different path, adopting a pro-innovation, principles-based framework that empowers existing sectoral regulators (FCA, Ofcom, CMA, ICO) to apply AI-specific guidance within their domains rather than creating a single AI regulator. For firms operating in both the UK and EU, navigating these divergent approaches is a growing compliance challenge.
Large Language Model (LLM)
An AI system trained on vast text datasets to generate, summarise, and analyse human language — the technology behind tools like ChatGPT and legal AI assistants.
Algorithmic Bias
Systematic errors in AI decision-making that produce unfair outcomes for particular groups, often reflecting biases present in training data.
Explainability
The degree to which the internal logic of an AI model can be understood and communicated to humans — a key requirement for high-risk AI under many regulatory frameworks.
Training Data
The dataset used to teach an AI model to recognise patterns and generate outputs — its quality and composition directly determine the model's capabilities and biases.
Model Risk
The risk of adverse consequences arising from decisions based on AI or statistical models that are incorrect, misused, or inadequately understood.
AI Governance
The internal policies, processes, and controls an organisation puts in place to manage the development, procurement, and use of AI systems responsibly.
Deepfake
Synthetic media — typically video or audio — generated by AI to convincingly depict events that did not occur, raising concerns in fraud, evidence, and defamation.