CIArb Guidelines on AI: Key Soft Law Step in Arbitration – Natalia Chumak

ABSTRACT

This article examines the evolving landscape of soft law regulating artificial intelligence (AI) in arbitration, with a particular focus on the CIArb Guideline and its practical implications. It highlights the challenges embedded in some of CIArb's recommendations and advocates for greater transparency and data collection regarding the use of AI in arbitral proceedings to better inform future policy and practice. The arbitration market more broadly needs to keep pace with rapid developments in AI, so as to, in due course, create a more uniform approach to the application of AI in arbitration globally.

I. INTRODUCTION

Guidance on the use of AI in disputes continues to flourish. Launched in March 2025, CIArb's Guideline on the Use of AI in Arbitration² is the most comprehensive addition to date. So, what does it offer?

The first dedicated soft law guidance in the international arbitration context emerged in November 2023, when the American Arbitration Association's International Centre for Dispute Resolution published its Principles Supporting the Use of AI in Alternative Dispute Resolution.³ In April 2024, the Silicon Valley Arbitration & Mediation Center (SVAMC) also published AI Guidelines,⁴ followed by the Stockholm Chamber of Commerce (SCC) Arbitration Institute last October.⁵ Viena International Arbitral Centre (VIAC) published its Note on the use of Artificial Intelligence in arbitration proceedings⁶ in April 2025 - after CIArb published its Guideline.

Whilst not directly relevant to arbitration, it is worth noting that some courts have also issued AI guidance, thus confirming the trend for leading judicial and arbitral authorities to provide much-needed thought leadership on the AI use in commercial disputes. Singapore Court introduced a Guide on the use of Generative Artificial Intelligence tools by court users in October 2024.⁷ In the UK, Courts and Tribunals Judiciary initially published Artificial Intelligence (AI) Guidance for Judicial Office Holders in December 2023⁸ with a revised version published in April 2025.⁹

They are all useful documents, but they provide a principles-based framework. But as AI proliferates, parties and arbitrators will need more detail. Delivering practical guidance, the CIArb AI Guideline takes that next step.

II. ANALYSIS

The Guideline comprises an Introduction, four substantive parts and two appendices, as follows:

• Introduction
• Part I: Benefits and risks of the use of AI in arbitration
• Part II: Recommendations on the use of AI in arbitration
• Part III: Arbitrators' powers to give directions / make rulings on the use of AI by parties in arbitration
• Part IV: Use of AI in arbitration by arbitrators
• Appendix A: Template Agreement on the Use of AI in Arbitration
• Appendix B: Template Procedural Order on the Use of AI in Arbitration

Parts I and II provide a useful foundation for more specific guidance in Parts III and IV. Parts III and IV, together with the Appendices, are the most innovative sections of the Guideline.

A. Introduction

The Guideline provides a useful vocabulary of AI-related terms. Most notably, unlike many of the above mentioned soft law instruments on AI, the CIArb AI Guideline does not create its own definition of Artificial Intelligence, adopting the OECD's definition instead. The definition adopted by the OECD is as follows:

AI is a machine-based system that, for explicit or implicit objectives, infers from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.¹⁰

B. Part I – Benefits and Risks of the Use of AI in Arbitration

Part I of the Guideline provides a comprehensive summary of how AI may affect arbitral proceedings.

1. Benefits

It begins by outlining the potential benefits (paragraphs 1.1-1.10). However, the list of benefits appears to conflate three different things - benefits of AI, criteria for evaluating the said benefits and examples of uses of AI with reference to these criteria.

The overriding benefits are improved efficiency and quality (paragraph 1.1). The criteria for evaluating improved efficiency and quality include:

1. increased speed in performing such workflows as data analysis (paragraph 1.3);
2. cost savings in relation to such workflows as production of hearing transcripts (paragraph 1.7);
3. enhanced search capabilities for legal research (paragraph 1.2)
4. improved consistency and structure for collecting and processing evidence and translations (paragraphs 1.5 and 1.6)
5. improved depth in analysing evidence (eg. identifying patterns and disparities between witness statements) and assessing the merits of the case (eg. predictions of potential case outcomes or on the likely response to certain arguments) (paragraphs 1.3 and 1.9)

Another benefit mentioned is equal treatment of parties (paragraph 1.10). The Guideline states that AI can be used to help remedy the "inequality of arms". This proposition is conditional on whether 'the right tools are available to an under resourced party at reasonable cost'. This benefit may materialise in future, once the legal AI market has matured and the use of AI tools has become widespread in arbitrations, both of which should make AI tools more affordable for an 'under resourced party'. Conversely, tech advancements will continue to develop at a high speed thus continuing to give better resourced parties a competitive edge. The Guideline does identify this as a risk, which is acknowledged in the "Risks" section of Part I.

2. Risks

The risks associated with the use of AI are set out in paragraphs 2.1 – 2.9. Paragraph 2.1 summarises the key risks which include challenges to arbitral awards at the enforcement stage and adverse implications for fundamental due process rights. Such rights include the right to present a party's case and the impartiality and independence of the tribunal.

Other risks posed by AI use include loss of confidentiality, data security concerns, and over-reliance on systems that may produce unreliable or unverifiable outputs, known generally as 'hallucinations'.

The CIArb Guideline also identifies a wider and more global risk created by AI tools, or rather, the infrastructure needed to support the creation and maintenance of such tools. It is a risk to the environment due to the high levels of energy consumption by the infrastructure required to support AI.

C. Part II – General Recommendations

Part II sets out general recommendations for all participants in arbitration proceedings, including arbitrators, parties, and counsel. These recommendations flow from the list of benefits and risks set out in Part I.

The spirit of this section is to encourage participants to be diligent in integrating AI tools into arbitration. Importantly, it also serves as a reminder that the use of AI tools does not reduce the participants' level of responsibility and accountability.

Key points include:

• Understanding AI Tools: Parties and arbitrators are encouraged to make "reasonable" efforts to understand the technology, functionality, and data behind any AI Tool used in the arbitration (paragraph 3.1). The concern here is that there will undoubtedly be a variation in how "reasonable" is interpreted.
• Assessing Risks v. Benefits: Parties and arbitrators are encouraged to assess the potential risks and benefits of AI use (paragraph 3.2).
• Knowledge of AI-related law: Parties and arbitrators are encouraged to make "reasonable enquiries" about applicable AI-related laws, regulations, or court rules in the relevant jurisdiction (paragraph 3.3). In practice, this is likely to require considering AI-related regulation in the relevant jurisdictions at a very early stage of the proceedings and involving legal counsel from the relevant jurisdiction/s to assist with the review.
• Responsibility and Accountability: As mentioned above, the Guideline makes clear that the use of AI does not reduce a participant's accountability. Unless expressly agreed otherwise, all parties remain fully responsible for their actions and submissions, regardless of whether AI tools were involved (paragraph 3.4). In the light of recurring AI-related scandals in the press - which are primarily US-based - concerning the lack of verification of AI-generated material by lawyers, this statement is a useful reminder that AI tools do not change professional standards.

D. Part III – Parties' Use of AI in Arbitration

Notably, Part III encourages arbitrators to be proactive in regulating the use of AI in arbitration proceedings including:

1. appointing experts who understand relevant AI tools and their consequences (paragraph 4.2);
2. requiring disclosure of AI use (paragraph 4.4 and section 7)
3. recording decisions on the use of AI in a procedural order and, if the use of AI was a contentious issue, in its award (paragraph 4.6);
4. taking appropriate steps to remedy a party's failure to comply with the arbitrator's directions on the use of AI, such as making further rulings, awarding costs, and drawing adverse inferences (paragraph 4.7)
5. addressing the use of AI in the proceedings from the outset (paragraphs 5.2 and 5.3), which includes:
   (a) Reviewing a request for arbitration to determine whether and how the parties provided for the use of AI in their arbitration agreement; and
   (b) Inviting the parties to address the use of AI at the first case management conference or later.

E. Part III – Specific Provisions on Arbitrators' Powers to Rule on Admissibility of AI-Generated Material

On admissibility, arbitrators may be requested by the parties to make a ruling on the use of AI (paragraph 6.1).

Paragraph 6.2 provides 'If the arbitrators consider that the use (or non-use) of AI by one or more party jeopardises the integrity of the arbitral proceedings, arbitrators may make a ruling on the use of AI of their own motion after consulting the parties.'

Paragraphs 6.4 – 6.6 identify issues that arbitrators may consider or information they might request from parties which enable a ruling on AI use to be made.

But these provisions assume that arbitrators have significant AI literacy and imply that parties have sufficient information about relevant AI tools, even though some tech companies lack transparency in providing it.

Under paragraphs 6.7–6.8, arbitrators must consider applicable AI-related laws, including: a) the law of the arbitration seat; b) laws and rules governing proceedings; c) national laws of the parties; and d) applicable ethical rules. Although they may not directly regulate AI, such laws may be indirectly relevant to AI usage in arbitration.

Invariably, arbitrators will not always have sufficient expertise in AI-related laws to determine their effect in practice. They will therefore require parties to make submissions and/or provide appropriate expert evidence. This is likely to add time and cost to the arbitration process in some cases which may negate the efficiencies achieved from the use of AI tools in the first place.

F. Part III – Specific Provisions on Arbitrators' Powers to Rule on Disclosure of the Use of an AI Tool

Section 7 outlines parameters where disclosure may be required, encouraging arbitrators to be proactive. Paragraph 7.3 gives arbitrators powers 'Unless otherwise agreed by the parties, and after consulting them, … to impose certain AI-related disclosure obligations on the parties…'

This is another proactive step. In comparison, although SVAMC AI Guideline 3 states that disclosure is unnecessary, the commentary outlines when it may be appropriate. But there is no suggestion that arbitrators should take a proactive role.

G. Part IV – Use of AI in Arbitrations by Arbitrators

Like the SVAMC AI Guidelines, the CIArb AI Guideline encourage arbitrators to consult with parties on AI tool usage and provide them with an opportunity to comment. They further state that arbitrators "should" refrain from using the specified AI Tool, although this stops short of obliging them to do so if the parties disagree.

III. APPENDICES

Appendix A contains a template on the use of AI in arbitration which can serve either as a standalone agreement or as part of an arbitration agreement. If incorporated into the substantive contract between the parties, potentially years before any dispute arises, the approach adopted in the agreement on the use of AI may become obsolete.

Appendix B contains short and long versions of a procedural order on the use of AI in arbitration.

Even if not adopted in their original form, the templates are a useful starting point for parties and arbitrators: more than other guidelines, they are a step in the right direction.

IV. CONCLUSION

The CIArb AI Guideline is a helpful tool which aims to ensure that parties and arbitrators are aware of the relevant issues, actively address them in the arbitration process, and if they are unable to agree, ensure that the arbitrators can rule on them.

The CIArb AI Guideline also identifies some challenges which must be addressed by the arbitration community.

First, all stakeholders in the arbitration process should achieve a high level of AI literacy. The credibility and integrity of the process depends on their ability to stress-test AI tools and AI-generated information.

Second, costs and delays may increase in some proceedings as a result of expert evidence on AI-related issues and clarifications on applicable mandatory AI-related laws. These will probably reduce once practices become more established.

Third, and more generally, the soft law approach to regulating the use of AI is likely to create inconsistency in interpretation and application. Disagreements that require a ruling and place a heavy burden on arbitrators: any ruling which improperly permits, or restricts, the use of AI could still lead to an award being challenged. Given that arbitral awards are typically confidential, soft law will take time to evolve into reliable, authoritative guidance that is uniformly applied.

Finally, AI tools are developing fast. Even though practitioners are working hard to keep pace, some commentators suggest that guidance soon becomes outdated. AI's rapid evolution in international arbitration therefore makes it imperative for industry bodies, arbitral institutions and practitioners to respond quickly to developments. In addition to the need for a prompt response, a collaborative approach is required in order to address AI challenges and opportunities, to preserve credibility of the arbitral process, and to reduce the risk of challenges to arbitral awards arising from the use of AI.

Several leading institutions have commenced work on creating further guidance on the use of AI tools. These guidelines are expected to add depth to existing soft law regulation. For example, in April this year the ICC Commission on Arbitration and ADR unveiled its new Task Force on AI in International Dispute Resolution which will work on creating guidance on the use of AI.¹¹ The Civil Justice Council also established a new working group on the use of AI in June this year.¹²

Going forward, it would be very helpful if annual reports published by arbitral institutions began including statistics on the use of AI in arbitral proceedings. In particular:

1. the number and types of cases where the use of AI tools was addressed by the parties and/or arbitrators in PO1 and, separately, in awards;
2. the number and types of cases where there were awards on costs relating to the use of AI tools;
3. the number and types of cases where disclosure requests or orders were made relating to the use of AI tools;
4. the number and types of cases where awards were challenged on the grounds relating to the use of AI tools.