As AI tools have become more sophisticated, and more prominent in many areas of life, the legal profession continues to wrestle with the way in which it can be used sensibly and ethically in legal practice.
AI does not have the ability for legal reasoning, or critical thinking for decision making. In whatever capacity AI is used in legal work, legal analysis should be outside of the scope of a chatbot.
But there are ways in which AI can improve efficiencies in the process of international arbitration. While there is no formal regulation on the way in which AI is used in international arbitration, several useful guidelines and research have been produced to help lawyers navigate its use:
-
CIArb Guideline on the use of AI in Arbitration (2025) CIArb Guideline’)
The Chartered Institute of Arbitrators Guideline on the Use of AI in Arbitration (2025) is guidance, which purports to helps parties and their legal representatives to take advantage of the benefits of AI, and mitigate some of the risk to the integrity of the arbitration process.
It includes a precedent agreement on the use of AI in arbitration, which parties can use at the outset of their arbitration. It also gives parties a draft a Procedural Order on the use of AI in arbitration.
-
Silicon Valley Arbitration and Mediation Center: Guidelines on the use of Artificial Intelligence in Arbitration – 31 August 2023 (The ‘SVAMC Guidelines’)
The SVAMC Guidelines are a “principal-based framework” for the use of AI tools in arbitration. The guidelines help participants to navigate the potential applications of AI.
These guidelines also provide a Model Clause that can be incorporated into Procedural Orders to make the Guidelines applicable to all participants involved in a particular arbitration proceeding.
-
2025 International Arbitration Survey – The path forward: realities and opportunities in arbitration (2025 Survey Report)
White & Case partnered with the School of International Arbitration at Queen Mary, University of London to prepare this research, based on a survey of the international arbitration community. There were 2,402 questionnaire responses and 117 interviews conducted in order to compile this report.
The research explains that the use of AI is expected to grow significantly over the next five years, driven by the potential for efficiencies. However significant concerns persist about accuracy, ethical issues and AI’s ability to handle complex legal reasoning.
In this article we look at the ways in which AI could be used in international arbitration, and consider whether or not each of these potential applications is a viable use. We consider the capacity and limitations of current AI tools, and on the guidance that has been produced so far for application within the industry.
How can AI be used in arbitration?
The most accepted use of AI at the moment is as a tool to speed up labour-intensive tasks.
e-Disclosure
Technology-assisted review (TAR) helps legal teams to review vast amounts of documents and find the most relevant documents in a fraction of the time. Large Language Models and Machine Learning can find relevant documents based on responsive keywords. As human reviewers tag the documents, the platforms learn which documents are most likely to be relevant.
This gives the legal teams the most highly relevant documents near the start of the review. It saves time and costs as the legal teams may agree (at some stage) to take an approach where they do not review documents that are unlikely to be relevant, based on the agreed keywords, and the information the platform has learned within the review so far.
The SVAMC Guidelines lists as a compliant use “using AI tools to identify and select the documents potentially relevant and responsive to document production requests while disclosing the manner in which such tool was used in a way that would permit the opposing party to make an informed objection.”
Preparing chronologies and compiling procedural histories
It may be possible for AI tools to extract information from documents which allow it to prepare a chronology or events, and procedural histories.
However, at the start of a dispute, the facts are often unclear. It could be easy for legal teams to miss key facts about the case, if they are solely relying on AI to extract the relevant issues for the chronology.
If the key facts are already known, AI can be put to tasks to find the documents that evidence those key facts, and compile a chronology.
Finding arbitrators
SVAMC Guideline lists as a compliant use of AI: “using a specialised AI tools to conduct research on potential arbitrators or experts for a case, being mindful of the AI tools’ limitations and evaluating the results accordingly.”
A tool can analyse the experience of arbitrators as stated in their CVs to find a suitable list of arbitrators more quickly. Of course, this relies on the quality of the input. Practitioners may want to limit the field to arbitrators associated with certain chambers, or within a certain jurisdiction. There will also be a need for human input to assess the selection critically.
But AI could save some time in preparing an initial list of possible arbitrators for the dispute.
Transcription of hearings
AI note-takers are fairly sophisticated now. They can take verbatim notes of hearings, and provide a succinct summary.
However, there is limited practical use for transcription of hearings in the context of international arbitration. Even if transcription were required, would it be more helpful and accurate to have a video recording instead?
Use with caution: Tasks AI may be able to carry out with human oversight
AI is a supportive tool, but in most circumstances, its output needs to be checked. This is particularly important in the context of legal work. AI does not understand the law, and will mischaracterise the legal position, or the nuance in the law.
Stating the obvious, AI lacks cognitive thinking. Its output is based on statistical likelihood derived from training data. It is no replacement for analytical legal minds who can logically work through legal reasoning.
With that caveat firmly in place, there are a few tasks that AI could make more efficient, so long as its output is reviewed analytically.
Drafting correspondence
The 2025 Survey Report explains that some arbitral institution staff found that AI was highly effective at preparing and formatting standard correspondence.
In the context of inter-parties correspondence, AI may be helpful in creating a first draft. But its use in this regard is limited. For factual correspondence such as arranging meetings, or mediation, it may serve a purpose.
The biggest criticism of AI in drafting correspondence is that it lacks the nuance to create the right tone. One respondent in the 2025 Survey Report describes AI-generated drafts as “too bombastic.”
Perhaps the time saved in preparing a first draft is eradicated by the time spent in correcting and amending the output.
Translation and interpretation
Certain plug-ins can automatically translate multilingual e-disclosure data held on Relativity and other e-discovery platforms. This can be an excellent way of reviewing documents in different languages in complex cross-border arbitrations.
When it comes to the translation of legal documents however, human translators with legal experience are superior to the machines. Human translators capture the meaning of the text, and they’re able to articulate legal particularities that may exist in one jurisdiction but not another.
Tasks to avoid with AI
The tasks below have been mooted as potential uses for AI, but they appear to be beyond the capabilities of current AI tools. Using AI in this way could present unnecessary risk to the professional integrity of legal counsel, and, in the most egregious examples, it could even jeopardise access to justice.
Drafting submissions
In the 2025 Survey Report quality 72% of respondents reported never using AI for drafting submissions, citing concerns about accuracy and reasoning.
The Report says that “in interviews, junior counsel and institution staff appeared more inclined to use AI for first drafts, whereas more seasoned counsel and arbitrators were more resistant, citing quality control, reputational risk or a wholesale rejection of delegating tasks requiring human judgment to AI tools.”
The problem is that AI cannot carry out the legal reasoning necessary for a coherent, persuasive submission.
However, the SVAMC Guideline does not rule it out, and instead lists as a compliant use:“Using AI tools to assist with drafting language for pleadings or written submissions where the final work product is fully source-checked and vetted for accuracy from a factual and legal standpoint.”
This appears to say that AI can do a first draft, so long as trained legal counsel reviews it. But, given the level of thinking and analysis required to write a submission, the output will always be more coherent when it is written from scratch from counsel who are experts in putting together arguments and legal reasoning.
Evaluating legal arguments
Can AI evaluate legal arguments?
It will probably tell you that it can. But the reality is that it will oversimplify the arguments, and discard the nuance. It will confidently deliver a half-baked summary, which lacks analysis. It cannot competently interpret the facts and the law and it cannot apply the law to the facts of the specific circumstances.
Factual and Legal research
AI tools like Chat GPT are known to make up (or ‘hallucinate’) case law. Lawyers will have to check every case that Chat GPT suggests to find out if it exists, and if Chat GPT’s interpretation and analysis of it is correct. In reality, this creates more work than any other avenues of legal research.
Summarising witness statements and other documents
While AI has the capability to provide succinct summaries of documents, confidentiality may be an issue, if open source AI is used for this task. Confidential information should never be entered into open source AI, as the software uses its input as learning. Confidential information could then be used as answers to another user’s question.
Summary
There is a healthy scepticism about the use of AI, even as it becomes more prevalent in different industries and areas of work. For international arbitration, AI can be used to make document review and disclosure more efficient. However, it is a poor replacement for any tasks that require legal analysis. It is only effective as a tool for labour-intensive tasks but cannot be trusted as a standalone authority, without detail human oversight.
