Signature and TransPerfect Legal – AI in Arbitration: Key Takeaways from our panel session at PAW 2025

Our Panel of Experts comprised: Dr. Maxi Scherer (ArbBoutique), Bernhard Maier, Knud Jacob Knudsen (Simonsen Vogt Wiig) and Al-Karim Makhani (TransPerfect), and the discussion was eloquently moderated by Phillip Kurek an arbitration Partner at Signature.

Defining AI in the context of arbitration

First off, a simple definition of AI. Al-Karim Makhani explained, amongst many complicated and scientific definitions, a straightforward way to understand AI is that "It tries to simulate human intelligence by using large amounts of data and processing power to predict outcomes." Simulation of human cognition as opposed to independent human thought became an interesting area of debate for the panel.

AI has already become embedded in arbitration through early applications like speech-to-text transcription during hearings, machine translations, and sophisticated eDiscovery tools that use predictive coding to streamline document reviews.

These technologies significantly increase efficiency by automating routine tasks, allowing arbitration professionals to focus on more complex advisory work. Panellists agreed that, while AI is not capable of replacing human judgement, it considerably enhances the quality, speed, and consistency of legal work.

Emerging Technologies and Their Impact

Dr. Maxi Scherer highlighted two advanced AI models that are particularly impactful in arbitration contexts: reasoning models and retrieval-augmented generation (RAG). "These models are 'thinking' longer before they give you an answer," Scherer said, explaining how these models perform deeper analytical tasks using more comprehensive automated prompts within the model itself. The RAG model combines a ChatGPT-style large language model with dedicated legal databases, significantly improving the accuracy and reliability of generated information.

This technology mitigates common issues with traditional AI systems, particularly "hallucinations," where AI generates plausible but incorrect information. By cross-referencing outputs with verified legal resources, RAG models greatly improve and importantly allow the audit of AI output.

One of the highlights of the panel was Dr. Scherer’s summary of a recent study on the subject. A recent empirical study by two U.S. law professors involved 120 law students completing six common legal tasks—such as drafting client letters, research memos, and NDAs—with half the group using AI tools and the other half working without them.

The AI-assisted group consistently produced work that was faster, clearer, and more professionally drafted. Notably, their legal analysis was also stronger in three of the six tasks, with mixed results in two, and underperformance in only one task—drafting an NDA—possibly because it is a simpler task that humans can already handle efficiently by using templates. The study highlights that the real power lies in collaboration between human and artificial intelligence, suggesting that AI can enhance legal work, but critical human oversight remains essential.

AI's Potential in Practice

Knud Jacob Knudsen provided practical insights, illustrating how practitioners actively employ AI in their daily tasks: "I use AI for instance for translations… I use AI to improve my language," he explained. Additionally, Knudsen emphasised the usefulness of AI-generated chronological overviews, which simplify complex case timelines, aiding in quicker understanding and decision-making during arbitration proceedings.

Al-Karim Makhani added another practical application by highlighting generative AI’s role in document review: "You can draft these [prompts] from looking at pleadings, you can draft them from review protocols, and you can just run that across the set of data." These prompts facilitate rapid and accurate extraction of relevant information from large data sets, drastically reducing time spent on initial document reviews.

Risks and Cautions

Despite significant advantages, the panel warned against several inherent risks in AI use. Bernhard Maier cautioned that AI systems inherently lack awareness of their limitations: "These algorithms do not know what they do not know… they don't know when they don't actually know the answer." This limitation could lead practitioners to rely on incorrect or incomplete data, resulting in substantial legal errors.

Knudsen reinforced this point by referring to Mata v. Avianca, where reliance on unverified AI-generated legal documents led to serious professional consequences. "I think that one of the number one risks is not knowing what you're doing," Knudsen emphasised, advocating for thorough training and awareness of AI’s limitations among arbitration professionals.

Regulation and Responsible Use

The panellists highlighted the urgent need for appropriate regulations and guidelines governing AI in arbitration. Dr. Scherer noted, "The EU AI Act actually does apply to international arbitration," categorising AI use by arbitrators as "high-risk," thereby demanding stringent compliance.

The establishment of various soft-law frameworks, such as those developed by the Chartered Institute of Arbitrators (CIArb) and the Stockholm Chamber of Commerce (SCC), provides practitioners with practical guidance and templates for procedural agreements regarding AI use, significantly enhancing transparency and compliance.

Practical Recommendations

Al-Karim recommended a few practical guardrails to help arbitration practitioners use tools more responsibly. The starting point must be good old common sense - AI is a tool, not a substitute for legal judgment. Lawyers must always verify outputs, especially sources, and never submit AI-generated content without checking independently against primary materials. Confidentiality remains critical: public AI tools should not be used for sensitive, confidential or privileged information, while enterprise models offer secure, closed environments. Privilege is more complex, particularly with generative tasks that don't yet reflect a legal trend; protections should include clear engagement terms covering AI-generated work as part of a firm’s working papers. Firms should keep clear records of AI use, consider disclosing its application, and ensure someone in the team is responsible for monitoring the fast-evolving regulatory landscape.

Bernhard Maier further emphasised the need for comprehensive legal education about AI, recommending practitioners "be passionate about the ones and zeros” to “understand the technology underlying [AI]."

The panellists also stressed the importance of clearly documented AI use policies within arbitration practices, including methodologies for data verification, privilege protection, and confidentiality assurances.

Looking Ahead

Given AI’s constant evolution, panellists urged arbitration professionals to stay informed and adaptable. Knudsen summarised the sentiment, stating, "Dive into it… actively use it, and the more you use it actively, the more possibilities that you will see are there."

Dr. Scherer concurred, adding that the arbitration community must "engage… very enthusiastic about the use of AI because it necessarily can improve our output of work, but at the same time we also need to be very sceptical."

Bernhard’s key take away was that for lawyers to thrive in a digital legal world, they must go beyond the surface and develop genuine curiosity about the technology behind AI — understanding how it works is no longer optional, it’s essential.

Al-Karim summarised the clear consensus from the panel: AI is here to stay. Rather than a passing trend it will become a foundational part of professional and personal life. Soon, it will become so embedded in practice that we’ll stop talking about “AI” altogether. The key message for arbitration practitioners is to stay curious, use secure tools, and always maintain human oversight. While much of the current focus is on using AI to optimise existing processes, the real opportunity lies in reimagining how we work altogether. AI isn’t just a tool for efficiency — it’s a catalyst for creativity and transformation in legal practice.

As AI technology continues reshaping the landscape of international arbitration, the community must proactively engage in dialogue, regulation, and education, ensuring responsible and beneficial use of these powerful tools.