Regulatory Notice: UtterConnection is an independent operations consultancy, not an SRA or CLC regulated law firm.Read compliance info
Reference Guide

AI Glossary
for lawyers.

Every AI term your firm needs to know, explained in plain English. No computer science degree required.

Artificial Intelligence (AI)

The broad term for computer systems that can perform tasks typically requiring human intelligence, like understanding language, recognising patterns, or making decisions. In legal practice, AI usually refers to tools that help with document review, research, or drafting.

Large Language Model (LLM)

The technology behind tools like ChatGPT and Google Gemini. An LLM is trained on vast amounts of text and can generate human-like responses. Think of it as a very sophisticated autocomplete, it predicts the next word based on patterns, not understanding.

Generative AI

AI that creates new content, text, images, code, or audio, rather than just analysing existing data. ChatGPT, Microsoft Copilot, and Claude are all generative AI tools. For law firms, this is the type of AI most likely to cause both opportunities and risks.

Hallucination

When an AI produces information that sounds convincing but is factually wrong. In legal contexts, this includes fabricated case references, made-up statutes, or invented legal principles. The Upper Tribunal recorded a "considerable increase" in the latter half of 2025 in fictitious authorities cited in both statutory appeals and judicial review applications.

Prompt Engineering

The art of writing effective instructions for AI tools. A well-crafted prompt produces better, more relevant output. For solicitors, good prompting can mean the difference between a useful draft and a document full of errors. It's a skill, not magic, and it can be taught.

RAG (Retrieval-Augmented Generation)

A technique where the AI retrieves specific documents or data before generating its answer, rather than relying on its training data alone. This is critical for legal work, it means the AI can reference your actual case files rather than guessing. More accurate, less hallucination.

Open-Source vs Closed-Source AI

Two different meanings sit under these words, and mixing them up is how firms get this wrong. Technically, open-source means the model's code and weights are published so anyone can run them. In practice, and in the way the courts use the phrase, an "open-source AI tool" means one that is publicly accessible and where what you type may leave your control. ChatGPT's free tier is not open-source in the technical sense, but it is exactly what the Upper Tribunal meant.

The distinction that matters is where your data goes. In UK and R (on the application of Munir) v SSHD [2026] UKUT 81 (IAC) at [21], the tribunal put it directly: "Closed source AI tools which do not place information in the public domain, such as Microsoft Copilot, are available for tasks such as summarising without these risks." For client work, a closed tenant you control is the answer. Test the contract terms, not the marketing.

Token

The basic unit of text that AI processes. A token is roughly 3/4 of a word. When AI tools have "context limits" (e.g., 128K tokens), it means they can only process about 100,000 words at once. Relevant when uploading long legal documents for review.

Fine-tuning

The process of training an existing AI model on specialised data, such as your firm's precedents, house style, or specific practice area knowledge. Fine-tuned models can produce more relevant outputs but require significant data and expertise to build properly.

AI Agent

An AI system that can take autonomous actions, like sending emails, updating databases, or completing multi-step tasks without human intervention at each step. Powerful but risky in legal contexts where human oversight is essential. We recommend careful governance around any agent-based tools.

Automation vs AI

Automation: Rule-based, deterministic "if X happens, do Y." Predictable and reliable. AI: Pattern-based, probabilistic "given X, the most likely Y is…" Flexible but less predictable. The best legal workflows often combine both: automation for the routine, AI for the nuanced.

API (Application Programming Interface)

A way for different software systems to talk to each other. When we architect AI workflows and control systems for firms, we often use APIs to connect AI tools with your existing case management system, email, or document management, so everything works together without manual copying and pasting.

NLP (Natural Language Processing)

The branch of AI that deals with understanding and generating human language. It's what allows AI tools to read a contract and extract key dates, or to summarise a witness statement. Essential for legal AI applications.

Data Privacy

The rules and practices governing how personal and sensitive data is collected, stored, and processed. For law firms, this includes GDPR compliance, client confidentiality obligations, and ensuring AI tools don't retain or share client data inappropriately.

Encryption

The process of converting data into a coded format so it can only be read by authorised parties. Important when evaluating AI tools: you should verify whether your data is encrypted both "in transit" (being sent) and "at rest" (being stored). If a vendor can't answer this, walk away.

Want to go deeper?

Book a free AI Audit and we'll explain exactly which tools and techniques are relevant to your firm.

Start AI Reality Check →