Speech by The Master of the Rolls: Justice for all, justice for the accused
This speech was delivered by Sir Geoffrey Vos at the Old Bailey, London, on 4 February 2026, as part of the Justice For All series of events
Introduction
1. Many thanks to Alderman and Sheriff Robert Hughes-Penney for inviting me to deliver this keynote address.
2. I am told that it is intended that I will provide an “in-depth examination of the rights of the accused, delays in the Courts and justice systems, and the critical importance of procedural fairness and access to legal representation” and a consideration of “how technology and AI may improve the efficiency and fairness of the justice system”.
3. No doubt, one could address that ‘spec’ from a number of angles. Recent public debate around the future of our justice system has focussed on the proposed restriction of jury trials, the reduction in the availability of legal aid, the backlogs in the criminal courts and more. Those are matters for politicians, policymakers and others to resolve.
4. For my part, I believe that the solutions for the justice system of 2026 engage the fundamental principles under which it operates. It is, of course, axiomatic that we should be aiming to provide “Justice for the Accused” and, perhaps as importantly, justice for the businesses, individuals and families that are involved in some 15 million disputes, large and small, that arise every year within our Society in England and Wales.
5. To achieve those laudable goals, we need, in my view, to look for the solutions and opportunities that are offered by recent revolutionary technological advances. To grasp those opportunities, we need, I think, to look at the fundamentals of justice, rather than assuming that the ways in which we have been accustomed to delivering justice in the last 200 years is the only, or even the best, way of doing so in the second quarter of the 21st century.
6. So, with that introduction, I will consider the following issues briefly tonight:
(1) What are the fundamentals of delivering civil and criminal justice in the age of hugely capable AI – what I call the machine age?
(2) How can and should those fundamentals be delivered more quickly, more efficiently and at more proportionate cost, but still justly, in the forthcoming generation?
7. I should say at once that I am not suggesting that AI will solve all our problems, particularly where vulnerable people in the civil, family and criminal justice system, are concerned. But I do think that, in the context of the use of AI and technology more generally in the justice system, we have been guilty of some quite shortsighted thinking. It is important now to look beyond the short-term and to consider not only what machines can do to help deliver justice, but also what humans will still be required to do in the fully developed machine age.
8. To summarise, my thesis is that it is now perfectly clear that AI will be able to provide or assist in the provision of justice outcomes, whether civil or criminal, in a tiny fraction of the time that lawyers, judges and human actors require to do so. Economics may well eventually dictate that many of these technological solutions will have to be adopted. For that reason, our society needs to consider now, and urgently, what humans need to continue to be doing to preserve the fundamentals of justice for humans in that machine age.
9. I shall try to explain under the headings I have mentioned.
What are the fundamentals of the delivery of civil and criminal justice in the machine age?
10. The key to this question is as to the meaning of “justice”. It is a mistake to think that “justice” means exactly the same thing to the people of 2026 as it meant to the people of 1873 when our current civil justice system was enshrined in the Judicature Acts.
11. First, in 1873, justice was something that was generally reserved for those that could afford it. Criminal justice was rough and ready. There was little or no legal aid, and civil justice was simply not available to those without means.
12. In 2026, our citizens and businesses want to see a system where justice is available equally to the least privileged and the most vulnerable in our society as it is to the most wealthy and the most privileged. That is, of course, an aspiration that is hard to achieve, but it is important to understand that it is what our society expects. Providing justice for 60 million people from hugely diverse backgrounds is a far harder task than providing justice for the wealthy members of a non-diverse population of some 23 million people in 1873.
13. Secondly, in 1873, justice was delivered relatively quickly. Criminal trials were held soon after the arrest and sentences were executed with despatch.
14. The resolution of disputes quickly and efficiently is essential to the economic wellbeing of the country. Lengthy disputes affect the productivity and psychological wellbeing of those involved in them and cause significant economic drag.
15. Justice has become delayed in the years leading up to 2026 mostly because of the complexities of modern life, not because anyone thought that delay was to be preferred to despatch. Civil and criminal cases now involve massive quantities of data and documents that require to be processed and evaluated. All that takes time. Also, we have developed cumbersome rules of evidence that were suitable for former times, but may or may not be appropriate in 2026. Technology will enable us to rethink those rules of evidence so as to ensure that justice can be delivered more speedily with its assistance.
16. Thirdly, the sophistication of our citizens and communities in 2026 means that there are many more crimes and disputes per capita of population than there ever were in the past. Again, the complexity of our lives provides endless opportunities for wrongdoing and dispute, whether it is online scams, cyber-crime or road traffic offences. There may be less old-fashioned house burglaries than there were in yester-year, but there are many more disputes and minor infringements than there ever were. In our modern society, an ever-increasing section of our population expects the “authorities” to tackle each and every one of these justice issues.
17. A sub-set of this problem is the fact that AI is now being used by almost every individual litigant in person and small business. The first port of call used to be a lawyer if one was available and affordable. Now the first port of call is ChatGPT or CoPilot. Whatever answer generative AI gives, the would-be litigant in person can easily use it to transform a mass of documents and personal information into an arguable legal claim. This means that, in future, we will see many more civil claims because AI can create them free of charge, where previously claim numbers were limited by the availability and cost of lawyers. Judges and courts need to be ready to deal with an AI revolution that may vastly increase the number of civil, family and tribunals claims with which the justice system has to cope.
18. So, to summarise, the “justice” we are considering in 2026 must be available equally to all, and must be provided quickly and efficiently for many more people and situations than ever before. Victims are, of course, also important and no infringement or dispute can be ignored or left unresolved.
19. These are exacting parameters for justice, and start to explain why we find ourselves in the situation we are in. Changes in the expectations of our society occur imperceptibly, and there has, perhaps as a result, never been a sufficient rethink of what we actually want to achieve in the delivery of justice. The modern expectations of our society and citizens and businesses have simply developed and appeared without discussion or debate.
20. I shall not discuss tonight whether and how such a rethink would be appropriate. That would be a highly controversial subject. Some would, no doubt, argue that our society cannot afford to deliver justice for all. Others would argue that justice for all is an essential component of the rule of law and of a democratic society. I do not believe the debate is helpful, because it is a fact that justice for all, in the way I have described it, is what our society has come to expect, and it is the duty of any self-respecting justice system to deliver what the society it serves has come to expect. The justice system only exists to serve, and we judges have to be careful not to think that the system we operate has any function beyond selfless service to our society and communities.
21. It can now, I hope, be seen that the fundamentals of justice in 2026 are the delivery of speedy, efficient and cost-effective outcomes for the many civil and family disputes and criminal infringements that arise every year.
22. As judges and lawyers, and in fact politicians, we would, I think, be totally failing in our duty if we did not recognise that technology and AI can play a large part in delivering justice of the kind I have explained.
23. That brings me, then, to my second question.
How can and should the fundamentals of modern justice be delivered more quickly, more efficiently and at more proportionate cost, but still justly in the forthcoming generation?
24. One of the problems with providing a definitive answer to this question is that technology is moving very fast, and the thought leaders in our society have been struggling to keep up with its capabilities. If we were to devise solutions based on today’s AI capabilities, those solutions would be outdated in months.
25. The first assumption is and must be that AI will be far more capable than many can now imagine. It is pointless to dwell upon the hallucinations of today, when we know that such hallucinations will very likely soon be things of the past. We need, as I have said, to consider the more fundamental questions of how justice should be delivered when AI is able to decide cases, both civil and criminal, as or more reliably than humans and certainly far more cheaply and quickly. We are now not far off the age of quantum computing, which will radically change online and offline security, and allow for far more sophisticated verification and authentication as well as the reverse. That change, by itself, is likely to make a seismic change to our approach to evidence gathering and validation.
26. The second assumption that is difficult for some to accept is that it is not appropriate to allocate the same level of resources to the most minor disputes as to the most serious disputes. Perfection is impossible, and, actually, most people are content with a rougher and readier solution for a dispute about £50 than they are for a dispute about millions. As is often said, and in justice as elsewhere, the perfect can be the enemy of the good.
27. The third assumption that I would make is that individuals and businesses are likely actively to want their disputes, at least their minor disputes, decided by machines, once they understand that such resolutions will be far quicker and cheaper than waiting for human judges to decide outcomes using outdated analogue processes and procedures. Of course, there is the half-way house of judges taking decisions informed and assisted by AI. But, as I have said in other lectures, the problem with that approach is the time and expense that would be incurred by human actors checking and validating the advice given by the machine. It is also always possible to have machines decide at first instance with appeals to human judges, but this too raises problems about where the lines are to be drawn.
28. The fourth assumption that I would make, but with which not everyone will agree, is that humans need to be front and centre of final justice decisions affecting people’s lives. As I have also said elsewhere, I do not believe that it is by any means clear that a machine, however reliable, can be regarded as “an independent and impartial tribunal” within the meaning of article 6 of the European Convention on Human Rights (ECHR). Put another way, I do not believe that we should be building the delivery of justice solely on machine-made decisions, even if machine-made decisions must have a part to play in modern justice. Human judges must be central to final justice decisions affecting people’s lives, because our citizens will not accept a justice system provided for humans and delivered entirely by machines. This is not because humans will not necessarily accept some machine-made judicial decisions. It is because judges are the last resort for humans to have wrongs or alleged wrongs righted, and ultimately it must be a fundamental human right for humans to have final outcomes determined by other humans, at least unless they have accepted some other method of determination in advance.
29. Against that important background, I believe that we can and should be having an informed debate about how we deliver the kind of justice that our citizens and businesses demand, using new technologies and AI.
30. In that debate, we are not starting from scratch here in England and Wales.
31. We are not starting from scratch in the criminal field, where the Criminal Courts Improvement Group established in 2021 is well-ahead in considering these issues.
32. In civil, family and tribunals justice, we are already establishing a digital justice system under the auspices of the Online Procedure Rules Committee. The Digital Justice System creates an integrated and coherent framework of data standards and connections between pre-action dispute resolution platforms, legal information platforms and legal advice platforms. A single data set for every dispute will enable early resolution by mediation, ombuds determination, arbitration or, if necessary, transmission to the online court-based dispute resolution platforms that are already established.
33. The Digital Justice System has yet to make ubiquitous use of machine-made decisions, though some constituent parts undoubtedly do. Small disputes are, for example, resolved online by eBay; eBay promises to resolve such disputes within 48 hours of referral at zero cost. It claims to resolve 60 million such disputes every year – they are probably disputes about sales of small items or services. But this exemplifies why it is wrong to think, as I have already said, that in terms of effective just decision-making, one size fits all. It plainly does not. If millions of people are content to use eBay to resolve their disputes, the formal justice system should pay attention.
34. So, what might AI-enabled justice delivery look like?
35. It could, as it seems to me have the following elements.
36. First, a well-structured pre-action Digital Justice System aimed at providing reliable legal information and legal advice online and at resolving disputes online by mediation, ombuds processes and AI-suggested solutions.
37. Secondly, a new technology-enhanced approach to the verification, authentication and validation of evidence. The idea that we could still require every fact to be proved by human witnesses giving first hand oral or written evidence in the age of quantum computing is, to say the least, somewhat fanciful.
38. Thirdly, a consideration of how far it should be open to individuals and businesses to agree to machine-made judicial decision-making, in civil or criminal cases, and whether such agreement should be possible in interlocutory, final or even appellate contexts.
39. Fourthly, reconsideration of a number of statutory and ECHR issues that are relevant to European decision-making. These include: (i) a determination of whether or not a machine can properly be regarded as an “independent and impartial tribunal” for the purposes of article 6 of the ECHR, which provides that “everyone is entitled to a fair and public hearing within a reasonable time by an independent and impartial tribunal established by law”; and (ii) an understanding of the proper meaning of article 14 of the EU’s AI Act which provides that high risk-risk systems (including those connected with the administration of justice) “shall be designed and developed in such a way, including with appropriate human-machine interface tools, that they can be effectively overseen by natural persons …”. Even though the UK is no longer a member of the EU, article 14 could be problematic for EU investment in the UK, if UK justice systems were to violate it. As I alluded to earlier, the problem with requiring human judicial oversight of every aspect of the administration of justice is the time and expense of checking what the machine can do so much more quickly than a human.
40. Fifthly, assuming that the ECHR in some way entrenches the right of an individual or business to a human judicial decision (which may be a big assumption as I have indicated), there needs to be a proper evaluation of how far human decision-making can properly be informed by machine-generated advice as to how those decisions should be taken. That is both under article 14 of the EU’s AI Act and more generally.
41. In other words, it seems to me that the legal community needs quickly to address the questions I have posed to work out how AI can and should be used to expedite the delivery of civil and criminal justice, in the way that I have indicated is expected by our citizens and businesses. There is not a minute to lose.
Conclusions
42. Let me return to the “spec” for my talk. You will recall that I was to provide, in the context of “justice for all”, amongst other things: an in-depth examination of the … delays in the Courts and justice systems, and the critical importance of procedural fairness and access to legal representation” and a consideration of “how technology and AI may improve the efficiency and fairness of the justice system”.
43. My conclusions are that the very concept of justice has changed. It is now a firm and unshifting expectation that justice should be equally available to everyone in society, from the least to the most privileged. That expectation alongside the complexity of modern society and the existence of massive datasets creates many more cases at all levels, whether civil or criminal. That provides a challenge for the delivery of justice that can only be fulfilled by the adoption of new technologies and AI. Legal representation and procedural fairness have to be considered in that context.
44. A debate is necessary as to precisely how we use AI to deliver justice for all. All societies will need an expedited resolution of several knotty, even existential, questions:
- What judicial decisions should always be made by humans rather than machines?
- When, if ever, should an individual or a business be able to consent to a machine-made judicial decision?
- What rules of technologically verified evidence should be adopted in the age of ever-more capable AI and quantum computing?
- How do article 6 of the ECHR and article 14 of the EU’s AI Act affect these decisions?
45. I have no doubt that there will be enough for our Panel to discuss.
46. Thank you for listening.