The library
Every module stands on its own and is open to every reader. Search the library, or narrow it by topic and level, and each module page suggests what is worth reading first.
New to AI? Start with What AI is, and what it is not, the first module on the Beginner path.
Modules
16 modules
What AI is, and what it is not
The difference between rules written by people and patterns learned from examples, what “model” means, and where AI already sits in legal work.
Topics: AI basicsRegulation
How a machine learns from examples
How a computer program learns from marked examples, how developers test it on examples it has not seen, why it is sometimes confidently wrong, and why the examples shape what it learns.
Topics: AI basicsHow models work
Neural networks without the mathematics
How a neural network weighs each number it reads, with every weight pictured as a dial that training sets, and why the trained result is one large file of numbers.
Topics: AI basicsHow models work
How a chatbot writes an answer
Text cut into pieces, the next piece predicted one at a time, why the answer reads as confident whatever its accuracy, and how the hidden instructions and searched documents a chatbot adds shape its answer.
Topics: AI basicsHow models workUsing AI in practice
Does AI reason?
What happens when an AI model writes out its working before answering, where writing out steps helps and where it fails, and why models invent authorities, with a worked example from the Divisional Court.
Topics: How models workUsing AI in practiceLiability and professional duty
Using an AI assistant well
Writing a clear request, working in steps, keeping sources apart from the assistant’s own reasoning, checking every source, and what memory and saved instructions mean for client confidentiality.
Topics: Using AI in practiceAI basics
Where AI meets the law: a map
The points at which copyright, data protection, the AI Act and liability attach to an AI product, stage by stage, each question with its primary source and the Advanced module that treats it.
Topics: CopyrightData protectionRegulation
What a model is
The small amount of mathematics the course uses, what a trained model physically consists of, and how weights, biases and layers turn input numbers into an output.
Topics: How models work
How a model learns
How training measures a model’s errors and corrects them, why models generalise and memorise at once, and how the computing power used in training became a regulatory yardstick.
Topics: How models workRegulation
Words as numbers
How text is cut into short pieces called tokens, why the cutting causes odd failures, and how each piece becomes a point on a map of meanings.
Topics: How models work
The transformer
How a chatbot links words, stores facts and writes its answer word by word, and why long documents cost more and can be read less reliably.
Topics: How models work
From predictor to assistant
How a text predictor becomes an assistant: pretraining, scaling laws, post-training, reasoning models, mixtures of experts and distillation.
Topics: How models work
Sampling, hallucination and memorisation
How a model chooses each word, why it invents things, and what the research shows about models reproducing their training data.
Topics: How models workUsing AI in practiceCopyright
The system around the model
System prompts, retrieval, tools and agents, the payment protocols built for agents, and why prompt injection is so hard to fix.
Topics: Agents and systemsUsing AI in practice
Open weights, images and provenance
Fine-tuning, the licences for downloadable models and the AI Act’s open-source exemption, how image and voice generators work, and the techniques and EU duties for marking machine-made content.
Topics: How models workTesting and transparency
Looking inside and testing
What research can see inside a model so far, the duties to explain automated decisions under the GDPR, the UK GDPR and the AI Act, why benchmark scores mislead, and the safety frameworks behind releasing powerful models to vetted users first.
Topics: Testing and transparencyData protection
Guided paths
Each level is also an optional route through the library, in an order where each module builds on the one before. Readers who switch on progress tracking see how far along each path they are.
Switched on, it keeps a short record of the modules you open and the quizzes you finish in this browser’s own storage, and the record stays on this device.
Beginner path
For readers without a technical background, written without mathematics and with every term explained where it first appears.
Intermediate path
For readers who want the mechanics: the arithmetic inside a model, how training sets its numbers and how a language model writes.
Advanced path
For readers who advise on AI: the systems built around models, how they are tested, and the law of the European Union and the United Kingdom that applies to them.
- The system around the model
- Open weights, images and provenance
- Looking inside and testing
- Training data and copyright
- Personal data in models
- Regulating the technology
- Liability, agents and professional duty