AI Arbitration: Faster, Cheaper – But Are Parties Ready?

The message from Sir Geoffrey Vos, speaking at the London International Dispute Week event of “AI in Arbitration – Promises and Pitfalls” is clear: AI will reshape arbitration. The key question is not whether to engage with it, but how to do so safely and strategically. 

Sir Geoffrey Vos is the Master of the Rolls. Here we summarise some of the key points in his speech and provide our thoughts:

For commercial parties, AI has the potential to deliver decisions in minutes rather than months, and at a fraction of the cost. In an environment where arbitration has become increasingly complex and expensive, that proposition will resonate strongly with businesses focused on efficiency and cost control. Crucially, arbitration provides fertile ground for this shift. Unlike court proceedings, where legal and constitutional safeguards (including Article 6 ECHR) make fully automated decision-making difficult, arbitration is built on party autonomy. If parties agree to an AI-driven process, there is little, in principle, to prevent it. 

And yet, departing briefly from Vos’s speech, Courts around the world are signalling that there is a line between AI as an assistive tool and AI as a substitute decision-maker.

In Association des ressources intermédiaires d’hébergement du Québec (ARIHQ) c. Santé Québec (2026 QCCS 1360), the Quebec Superior Court set aside an arbitral award where AI had been used to produce the reasoning. As part of its reasoning, the court emphasised a fundamental requirement of procedural fairness: parties are entitled to have their dispute determined by the arbitrator they appointed. Where AI moves beyond a supporting role and effectively generates the reasoning, that principle is undermined and the validity of the arbitral award is put at risk.

Similarly, John LaPaglia v Valve Corporation, the claimant sought to vacate an award on the basis that the arbitrator had improperly delegated decision-making to AI, exceeding their powers under section 10(a)(4) of the US Federal Arbitration Act (FAA). The court dismissed the challenge on jurisdictional grounds without engaging with the substance. While inconclusive, this case, alongside the recent Canadian judgment, demonstrate that parties are already testing the limits of AI in arbitration proceedings.

Against that backdrop, Vos’s suggestion that hybrid models, where AI generates draft awards subject to human oversight, is persuasive. Such models seek to preserve due process and trust, while still capturing efficiency gains. But the direction of travel is clear: market pressures may drive parties towards greater automation, particularly where cost savings are significant.

The difficulty lies in enforcement and legitimacy. Even where parties agree to AI-driven processes, awards may face challenge on public policy grounds, for example under section 68(2) of the Arbitration Act 1996 or Article V(2)(b) of the New York Convention. The risk is particularly acute where the use of AI raises concerns about bias, opacity, or the integrity of the decision-making process.

Please see here for the full speech: Speech by the Master of the Rolls: AI in Arbitration – Promises and Pitfalls - Courts and Tribunals Judiciary.

If you would like to discuss how current or future regulations impact what you do with AI, please contact Tom Whittaker, Brian Wong, Lucy Pegler, Martin Cook, Liz Griffiths or any other member in our Technology team.  For the latest on AI law and regulation, see our blog and newsletter.