Speech by the Master of the Rolls: AI in Arbitration – Promises and Pitfalls
Cross JurisdictionalMaster of the RollsSir Geoffrey VosSpeeches
LIDW
BIICL and Jones Day
AI in Arbitration – Promises and Pitfalls
Jones Day offices, 21 Tudor St, London EC4Y 0DJ
4 June 2026 at 4.30pm
Sir Geoffrey Vos, Master of the Rolls
Speakers
Master of the Rolls, Lord Neuberger, Michelle Bradfield of Jones Day, Alex Chalk KC former Justice Secretary, Professor Eva Lein of BIICL new project on AI uses in arbitration, and Stuart Craft, Manging Director of FTI Consulting.
Introduction
1. I have spoken a great deal about how AI is changing the availability of legal advice and information and court-based dispute resolution. I firmly believe that it is inevitable that AI will eventually be used, not merely to facilitate judicial decision-making, but as the judicial decision-maker in some fields.
2. That is inevitable because AI decisions will be reliable, but cheaper and quicker than human judicial decisions. Surely, commercial parties would, in some cases at least, prefer a reliable machine-made decision made immediately at a fraction of the cost to a reliable human decision made at great expense many months or years down the line.
3. I have also suggested that getting from where we are now to where we might like or expect to be is not easy.
4. Court services in a democracy cannot just abandon human judicial decision-making in favour of machine-made judicial decisions. First, article 6 of the European Convention on Human Rights guarantees to everyone “a fair and public hearing within a reasonable time by an independent and impartial tribunal established by law”. I have expressed the view, with which some others disagree, that an AI can probably may not constitute an “independent and impartial tribunal” within article 6.
5. But whether or not that is right, and secondly, the right to a human judicial decision in crime, civil, family and tribunal cases is deeply entrenched in our constitutional settlement. It would require legislation to change it. So, however, attractive it might be for the parties to a personal injury claim to have their damages assessed in minutes by a capable AI, human judges cannot just be abandoned in the courts without legislative change. And such legislative change will take time. And there will be significant disagreement as to what should be changed and how. Should, for example, a final appeal to a human judge be retained in every case?
6. and, thirdly, the LLMs necessary to make decisions in particular kinds of commercial and other disputes have yet actually be created and tested to ensure their reliability and absence of bias.
7. These parameters are, however, rather different in the arbitration context. In arbitration, the parties are autonomous and they can agree the procedure they want to be used just as they can agree the law they would like to be applied. The parties can engage a human arbitrator or arbitrators or, if they wish, they can agree that their disputes will be finally determined by a machine. This latter process – using machines to decide arbitral disputes – has not yet become commonplace, but some 60 million people every year are content for their disputes on eBay to be finally determined by eBay within 24 hours, I very much doubt that those disputes are determined by humans. So, it is by no means far-fetched to envisage widespread AI Arbitration clauses.
8. Moreover, the AAA-ICDR (the American Arbitration Association – International Centre for Dispute Resolution) (AAA-ICDR) announced in December 2025 that it was offering an AI arbitrator for two party, document-only construction disputes. It is offering a “human in the loop” model, where the AI arbitrator will evaluate the merits of claims and prepare draft award, but a human arbitrator will review it to test the reasoning and evaluate the decision, in order to quote safeguard trust, transparency and due process.
9. So, what, if anything, is stopping 100% machine-made decisions in the context of arbitration? The answer is probably human circumspection and caution. But, in the end, money will probably prevail as it so often does, and parties will probably agree, in some cases at least, to go for the cheapest option, which is likely to be some form of AI-driven arbitration.
10. Should we be worried about such a development? In my view, not, not unduly. It is one thing to impose machine-made decisions on parties to court proceedings, who might justly claim a constitutional right to a decision made by a human judge and even an appeal to a human Court of Appeal. It is quite another thing to allow the parties to choose a process that expedites the decision-making to which they have agreed to be bound, in exactly the same way as we allow them to choose the applicable law and the jurisdictional forum for the resolution of their commercial disputes.
11. The bigger question, perhaps, is not whether such an AI process should be permitted, but what will happen when it goes wrong, as it surely will.
12. It would obviously be theoretically possible to try to challenge an arbitration award made in this jurisdiction on the basis that there was a serious irregularity under section 68(2) of the Arbitration Act 1996 in that: “the way in which [the award] was procured [was] contrary to public policy”, or to resist enforcement under the public policy provisions in Article V(2)(b) of the New York Convention.[1]
13. There is, though, an even more significant problem that the advent of machine-made arbitrators will highlight. That is the ever-increasing overcomplication of arbitral proceedings. This is something I observed when I regularly sat as an arbitrator before coming to the High Court Bench in 2009. I believe it is something that has increased, I believe it leads to increased costs for the parties and will help drive those same parties to a quicker and more cost-effective dispute resolution mechanism.
14. Personally, I think it is inevitable that AI will change the dispute resolution landscape dramatically. I am promoting, through the Online Procedure Rule Committee, an online Digital Justice System to resolve all civil, family and tribunal disputes quickly and efficiently, before court process begins, and at minimal cost. The same principles can and should be applied, at least in part, to commercial dispute resolution, and with the help of AI, they now certainly will. The European Law Institute is about to publish its guidance on Digitalisation of European Justice Systems.
15. I am not sure we can now predict accurately where this journey will end, but I am sure that lawyers will need to adapt quickly if they are to retain the trust of their commercial dispute resolution clients. Businesses will not pay for something they can get more cheaply from a machine. That applies as much to legal advice delivered by ChatGPT, Claude, Harvey or Legora, as it does to AI driven arbitrations.
16. In the first instance, though, the human caution and circumspection will mean that changes will be slow at first, but when they come, they may accelerate. I believe the legal community needs to prepare itself urgently.
Conclusions
17. Can I conclude with some brief, but important markers for our panel discussion.
18. First, AI and LLMs are not equal across jurisdictions. We are fortunate here in the UK in that US-based LLMs are, for us, relevant and comprehensive. That is not the case for many multiple language jurisdictions such as India, Africa and the Global South. AI should not be allowed to introduce greater inequalities into global commercial dispute resolution, caused by the forfeiture of digital and private law sovereignty to the dominant home states of LLMs such as the US and China.
19. Secondly, judicial decision-making is of a different order to decisions made by Governments affecting people’s rights and corporate rights (and consensual arbitral decision-making). Judicial decision-making is the last resort. Whilst, as I have said, it is inevitable that the resolution of small disputes at first instance will come to be facilitated by AI, it raises fundamental questions as to the rule of law if we are to deprive humans of final human judicial decisions as to their legal rights.
20. Thirdly, in all these debates about AI, do not forget one crucial feature. AI can make decisions and process data in seconds. But humans are still humans. And humans take days or months to process and reach acceptance of unwelcome outcomes. If AI reaches decisions affecting humans and their businesses super-quickly, we should not expect those outcomes to be accepted in the same time-frame.
[1] Article V(2)(b) of the Convention on the Recognition and Enforcement of Foreign Arbitral Awards (New York 1958).