Arbitration is an alternative means of dispute resolution to litigation proceedings in a court of law. It is often suggested that arbitration which is a private means of dispute resolution outside the court state system is cheaper and quicker than litigation. That is a matter of opinion and a matter of fact determined by the requirements of any particular dispute referred to it. As a lawyer arbitrator of many years’ experience it depends several factors including the subject matter, complexity of the case, the issues, the evidence, employment of experts and co-operation and attendance of witnesses. In arbitration the parties must pay for the arbitrator, the cost of the venue of the hearing, lawyers and experts fess and travel costs. Arbitration can be as costly in time and money as litigation and some and much of which depends on the conduct of the parties and their cooperation between themselves and the tribunal. Proceedings can be expedited by case conferences online, hearings online documents only arbitration and now of course with the implementation and usage of AI for certain purposes.
AI presents the possibility of more efficient and expeditious proceedings providing enormous possibilities in the area of procedural law. Whilst it has many advantages there is also some concern as to its regulation and control, accuracy and bias. From the lawyer’s point of view it can be advantageous in terms of dispensing with the reading of bundles of documents, summarising the evidence, preparing summaries of witness evidence, summarising and explaining technical documents, editing files of correspondence and drafting submissions that may be exchanged between the parties and by these means saving hours of fee earners time. AI can also dispense with what may be undertaken by secretaries or administrative staff in terms of tidying up documents inserting headings and pagination as well as reviewing the documents for duplication and sorting. AI can be useful in identifying people in its production of instantaneous transcripts of the evidence, and submissions as well as cross referencing documents to exhibits. It can also check for privileged documents inadmissible in evidence. Whilst these are all useful functions it must be remembered that AI is a digital technical artificial process with no human sense of intelligence or productive mental effect so it is vulnerable to mislead, misinterpret, or fail to detect that vital piece of evidence which can only be obtained by a thorough human forensic study.
In a recent article in Arbitration International (Vol.42, No.1) 2026 two Turkish international lawyers gave warning as to the problems that can arise by the practical use of AI in arbitration proceedings. They give the example of a construction case where AI would find difficulty in judging the nuances as to the tone of correspondence between disputing parties. It would fail to see how the relationship between parties for example the architect and the contractor’s relationship broke down over a period of time. This can be detected by lawyers with experience of these cases by reading through the whole of the correspondence and judging the tone of it. AI does not do so it may miss a point that could turn the case with useful cross examination.
As for arbitrators they are duty bound to resolve the dispute on the basis of their terms of appointment and terms of reference outlining the case by the parties, the applicable law and rules observing their contractual obligations to the parties under their terms of appointment. They need to make an award on the balance of probabilities on the cases as presented by the parties, to evaluate their arguments, analyse their submissions, test evidential credibility, weighing the evidence, and making reasoned decisions on the submissions in the context of the evidence in constructing their Award. To use AI to make an award is to risk passive acceptance of what the AI process produces. That is not what the parties want: they want the expertise of the arbitrator they chose. That point was made to me when I was teaching a group of experienced Chinese lawyers at Cambridge last year. I asked them whether they wanted to have an award made by a robot or by a human being. They said they would prefer an arbitrator, a human being, to make the award.
To guide arbitrators the Chartered Institute of Arbitrators has published its guidance on the use of AI in arbitration. It clearly demonstrates the risks to enforcement of awards, the rule of law, the administration of justice and credibility of the process itself.
Being a process of private dispute resolution confidentiality is a key element in arbitration. Questions as to protection of the information and freedom from cyber-attacks and protecting such information which is stored in some AI platform is of concern. Such information may contain vital data about the parties which if released or stolen would be damaging to the parties whether they were domestic or international corporations or others wishing to protect their reputations and secrets of their business.
Another potential problem is a particular risk in terms of potential algorithmic bias by AI selective processes. The question of affirmational authority and cognitive inertia may also arise. Thus, AI presents a risk to another vital element of arbitrators using it in terms of Independence and impartiality. Questions may also arise as to due process and fairness in how the parties are dealt with in the case. Is there equality of arms, are they on the same level playing field? What if one party uses AI and the other does not? To rely on AI predictions as to the accuracy of decisions is problematic in the absence of any human oversight.
It is therefore no surprise that in many countries they may restrict the use of AI which may pose practical problems for those practising in the international arbitration field.
At present it appears the parties are quite free to agree what artificial intelligence tools they will apply for arbitration. No doubt there will be different opinions on such usage. Just how a tribunal can ensure that the warnings about warnings given in the guidance by the CIA is absorbed as another matter. Questions as to professional indemnity insurance may need to be reviewed by arbitrators in such cases where they will be tempted to rely upon AI.
However, in a recent survey conducted by Queen Mary University of London, School of International Arbitration and White and Case, a leading international law firm, their survey concluded that 78% of their respondents agree that AI should be used more often in international arbitration although 68% admitted they never or rarely used it. It does seem that if some major law firms are using AI for numerous purposes, then others will follow. The survey pointed out that confidentiality remains a concern with Open Source AI. Tribunals have to be cautious. Whilst AI can boost efficiency it has its risks. There are obvious concerns about its accuracy it may be useful for a first draft of a submission, but counsel will need to be circumspect as to producing the quality required for submission to a tribunal. Quality control, reputational risk, and delegation of function to AI are questions that need to be addressed. In short, it can reduce the need for labour intensive tasks as suggested above and possibly save time and cost. It has that potential, but there must be a need as the survey suggested for clear frameworks and standard practises governing AI’s role in arbitration particularly as it states as regards transparency. Above all in terms of the arbitrator’s duty the survey rightly concluded that an arbitrator must know their case better than anyone else and AI cannot replace that fundamental duty.
Dr Michael Reynolds is a Chartered Arbitrator and Fellow of the Chartered Institute of Arbitrators. He is an alumnus of the School of International Arbitration, Queen Mary University and now teaches International Commercial Arbitration at the University of East London.
