ODEC 9410 - Large Language Models (LLMS) and the legal profession

to be defined

LLMs - think ChatGPT, Claude, Gemini, etc. - have moved astonishingly quickly from academic curiosity to the operating system of modern knowledge work. For future lawyers, this goes beyond “digital literacy”: it is a structural shift in how legal expertise is produced, signalled, and valued. This course examines LLMs as technical artefacts, as legal tools, and as socio-economic forces capable of reshaping the legal profession itself.

We study how these models work (data, embeddings, transformers, prompting, fine-tuning), what they can and cannot do (reasoning, hallucination, bias, security failures), and how they recombine with law’s own epistemic machinery: argumentation, authority, adjudication, legitimacy, and professional judgment.

Students learn to critically assess AI-generated legal analysis; experiment with prompt engineering, RAG systems, and agent architectures; analyse their implications for justice systems and legal labour markets; and confront the technical, ethical, and institutional challenges posed by increasingly capable models — including the possibility that parts of legal judgment itself may be automated.

This is a course about the legal profession’s future, taught with the tools that may define it.

Learning Outcomes.

By the end of the course, students will (hopefully) be able to:

Understand the technical foundations of modern LLMs (embeddings, transformers, inference, fine-tuning, reasoning models).

Evaluate LLM output with professional-grade epistemic discipline: when to trust, when to verify, and how to structure “epistemic hygiene” in practice.

Use LLMs effectively as research partners: drafting, argument scaffolding, retrieval-augmented research, and adversarial testing.

Analyse the impact of LLMs on the legal profession, including productivity, labour markets, and legal education.

Assess risks: hallucinations, bias, prompt injection, jailbreaking, over-automation, disempowerment, and systemic safety concerns.

Debate whether - and in which domains - AI systems could legitimately perform judicial, regulatory, or advisory functions.

Produce a rigorous case study of an LLM-enabled legal tool, organisation, or workflow.

A mini moot court/debate in groups of two during the semester on one of the predefined topics (65%),

A participation grade 35%