AI and the legal profession: what lawyers need to know now

Understanding AI: Professor Christoph Kletzer 

When it comes to understanding AI, Professor Christoph Kletzer argues that legal professionals should focus less on commentary and more on direct engagement.  

What resources should legal professionals prioritise to understand the impact of AI?

Fewer think-pieces, more primary sources. Two kinds are worth your time. First, the systems themselves: spend an hour a week actually working with ChatGPT, Claude or Gemini on real legal tasks until you have a feel for where they are reliable and where they confabulate. Second, a single honest technical primer over a hundred opinion columns. Stephen Wolfram’s What Is ChatGPT Doing? explains the mechanism without mathematics. Then read the instrument that will actually bind you, the EU AI Act, alongside the SRA and Bar Council guidance.

What skills should lawyers start developing in preparation for AI?

Two mindsets, both uncomfortable. The first is stochastic. AI buys enormous leverage at the price of certainty: it is right more often but guaranteed less often. Lawyers are trained to deliver certainty, so this cuts against the grain; the skill is to assess and price uncertainty rather than flee it, without surrendering the judgment that is your real contribution.

The second is managerial. Once a tool drafts in seconds, your value shifts from doing the work to directing and checking it: framing the task, supervising the output, owning the result. You become an editor, and a manager of machines.

What practical steps can law firms or legal organisations take to incorporate AI responsibly?

Start with three. First, decide clearly what may be put into a public model and what may not; client confidentiality does not survive being pasted into a chatbot. Second, keep a human signature on everything that leaves the firm, because responsibility cannot be handed to a machine that bears none. Third, run small, supervised pilots on low-stakes work, research memos and first drafts, and measure them honestly before scaling. Going slowly here is not caution for its own sake. It is how you avoid mistaking a tool's speed for its reliability.

Which area of AI literacy is most urgently missing in the legal profession?

An understanding of what these systems actually are. Lawyers tend to swing between treating them as oracles and dismissing them as toys, because they picture the model as either reasoning or guessing. What it does is stranger: it predicts plausible text from patterns, with no access to truth and no picture of the world. Once that sinks in, the rest follows. You see why it cites cases that were never decided, and why it sounds most confident exactly when it is most wrong. The gap is not technical knowledge. It is an accurate mental image of the thing.

How can legal professionals keep pace with AI developments without becoming overwhelmed?

Separate the noise from the signal. The weekly product announcements barely matter. The capabilities that actually change your work change slowly enough to follow. Pick two or three sources you trust, ignore the rest, and judge each new thing by one question: does it change what I can responsibly rely on? Usually it does not. Tracking the frontier is the vendors' job. Yours is to know what the tools on your own desk can and cannot do, which is a far smaller and more manageable thing.

As AI and technology can potentially blur traditional legal boundaries, what kinds of knowledge do lawyers need to engage with these challenges?

I would resist the premise a little. Technology rarely dissolves a legal boundary. It moves the facts the boundary was drawn around. A smart contract still has to be a contract; an automated decision still has to be attributable to someone who answers for it. So what lawyers need is not a new jurisprudence but an old discipline applied to unfamiliar facts: enough grasp of how a system really works, what a model or a ledger or an algorithm is actually doing, to see past the metaphors to the legal question underneath. The categories hold. The substrate changes.

What lawyers risk unlearning: Professor Sylvie Delaxroix 

Much of the conversation around AI focuses on what lawyers need to learn. Professor Sylvie Delacroix suggests that an equally important question is what the profession risks losing.

We tend to ask what lawyers need to learn in order to use AI well. The more searching question is what they risk unlearning. Legal judgment depends on a capacity that resists automation: recognising when a situation does not yet have a settled answer, and keeping it open long enough to think. Systems optimised for fluent, confident responses train us out of that habit, not by being wrong, but by making the act of questioning feel unnecessary.

It is tempting to think transparency answers this. It does not. Even where each individual output is interpretable, fully understood and explainable, a profession can still be hollowed out, because the loss is collective rather than individual: the shared practices of contestation and refinement quietly atrophy. Individual interpretability is wholly compatible with collective deskilling, and the second is the more serious risk.

The implications are practical. Treat AI outputs as interpretations to be interrogated rather than answers to be adopted. Protect the habits that keep juniors challenging a machine-drafted document. And insist that the profession help shape how these systems express uncertainty rather than inheriting design choices made elsewhere. That last point is the heart of what we are building at the Centre for Data Futures.

Approaching AI education from both a practical and regulatory perspective: Dr Petros Terzis 

Dr Petros Terzis highlights how The Dickson Poon School of Law is responding to the opportunities and challenges presented by AI, combining practical AI literacy with a deeper understanding of the legal, regulatory and policy questions surrounding emerging technologies. 

At The Dickson Poon School of Law, we are fortunate to build on the skills and expertise of various academics who are at the forefront of the rapidly evolving field of AI and Law. This includes people who work on the integration of AI tools in legal practice and the judiciary as well as those working in the regulation and governance of  AI around the world. Our teaching portfolio and education programs reflect this rich diversity.  

On the one hand, the Law School offers the ‘AI literacy for Law’ open course. Led by Professor Dan Hunter, the course pierces the veil of Generative AI and enables its participants to understand the technology to harness its potential and integrate it into their personal and organisational workflows, for boosting accuracy, productivity, and strategic advantage. Regardless of background, participants gain practical skills in using AI tools, a solid understanding of the ethical and strategic implications of AI adoption, and an understanding of how legal practice is evolving with the technology. Alongside that, our MSc in Law and Professional Practice course prepares non-law graduates for practice by integrating AI into classroom activities, teaching the next generation of lawyers how AI can be incredibly useful for tasks such as drafting. Students are taught to critically analyse the outcomes to ensure that they become discerning, yet effective, users of AI.  

At the same time, our LLB programme, which I lead, offers a module on Generative AI and Law that is dedicated to global regulatory and policy developments relating to the development and deployment of generative AI systems and models. The syllabus is heavily informed by the latest developments in research and technical practice. The module covers the legal implications of Generative AI for data protection, platform liability, copyright law, and digital markets. Students invest time (and intellectual energy!) in learning how Generative AI models are trained and function. This familiarisation happens through methodical guidance across the model’s architecture, accessible readings and a selection of the latest legal cases, policy documents, and research articles from various jurisdictions. Students learn how to inform their legal analysis with technical considerations on the model’s architecture.

When dissecting the ‘technicalities’, complex details are avoided (not everyone needs to learn how to code!) and for every detail we signify (from web search in large language models to model memorisation), students are constantly asked to reflect on the question: ‘Why may this technical detail be relevant for law and policy?’. This course, along with the more domain-general ‘Technology Regulation and Policy’ are suited for students who consider a career in the law, technology, and compliance sector.  

Through this diverse offering, our ambition is to support our students’ journey in becoming thought leaders and professionals, capable of interpreting, presaging, and responding to the recurring fluctuations of a rapidly changing world. 

Conclusion 

As AI continues to reshape legal practice, the profession must balance technological fluency with the judgment, accountability and critical thinking that remain at the heart of the law. At The Dickson Poon School of Law, these challenges are already informing teaching, research and professional education, equipping current and future lawyers with the skills and perspectives needed to engage with AI confidently and responsibly. Whether through responsible use of AI, thoughtful regulation or innovative legal education, the challenge for lawyers is not simply to adapt to technological change, but to help shape it.