AI Arbitrators: Ushering in a New Era of International Arbitration

In a recent episode of Law Next, host Bob Ambrogi spoke with Bridget Mary McCormack, President and CEO of the American Arbitration Association (AAA), and Diana Didia, its Executive Vice President and Chief Technology and Innovation Officer, about the AAA’s bold leap into AI-driven dispute resolution. The duo announced something unprecedented: the launch of an AI-powered arbitrator in November 2025. Unlike traditional AI tools that merely assist practitioners, this system will evaluate case merits, generate recommendations, and prepare draft awards, all under the oversight of human arbitrators who validate and sign off on final decisions. This development marks a significant milestone as AAA approaches its centennial, signaling a transformative shift toward an AI-native vision for dispute resolution.

The ICDR’s AI Arbitrator Initiative

The International Centre for Dispute Resolution (ICDR)—the international division of the AAA—confirmed that AI arbitrators will be available from November 2025 for eligible document-only construction disputes. While headquartered in the U.S., the ICDR administers cases globally, including in energy and construction sectors, and maintains a worldwide network of hearing facilities through cooperative agreements, including with the U.K.

The AAA-ICDR has consistently demonstrated forward-thinking leadership in integrating AI responsibly into arbitration. In October 2024, it introduced AAAiPanellist Search, an internal tool to help case managers identify suitable arbitrator candidates during the nomination process. The AI arbitrator represents the next step in this evolution.

How the AI Arbitrator Works

The ICDR’s AI arbitrator will not act autonomously. Instead, it functions as a co-pilot, leveraging a structured legal prompt library and conversational AI to draft awards, which are then reviewed and validated by human arbitrators. According to ICDR, the system has been trained on a dataset of over 1,500 annotated construction awards, ensuring high-quality reasoning and contextual accuracy.

Parties will also have an opportunity to validate the AI’s understanding of their submissions before any draft award is prepared, reinforcing transparency and fairness. As the ICDR’s executive vice president aptly described, this is a “very muscular co-pilot,” not a decision-maker.

Benefits and Implications

The anticipated benefits are substantial: cost reductions of 30–50% and faster resolution timelines. Given the rising costs and extended durations of international arbitration—illustrated by the LCIA’s 2024 analysis showing average case durations increasing from 16 to 20 months—these efficiencies could reshape dispute resolution dynamics. For industries like construction, where cash flow and timelines are critical, this innovation could shift the balance from compromise to resolution. Also, it is possible that given the reduced costs, there will be an increased willingness of parties to pursue claims rather than abandon them for financial reasons.

Some Practical Issues with AI in Arbitration

While AI arbitration brings efficiency, cost savings, and speed, it also raises important issues around human judgment, oversight and ethical obligations. One of the significant drawbacks would be the inability of AI to replicate the nuanced skills of arbitrators, such as encouraging mediation or exercising equitable discretion. The following are some of the key concerns I have regarding AI Arbitrators.

1. Lack of Mediation or Settlement Push: Human arbitrators often encourage parties to settle or mediate, using subtle persuasion and empathy. AI agents lack this interpersonal capacity, which may reduce opportunities for amicable resolution before an award is issued.
2. Limited Exercise of Discretion: Many arbitral outcomes involve equitable reasoning or context-sensitive judgment, which AI tools (trained mainly on precedents and structured data) may not adequately replicate.
3. Perception of Legitimacy: Parties may feel skeptical about the fairness of an “AI-driven award,” even when subject to human oversight, raising risks of challenges to enforceability.
4. Sector-Specific Expertise vs. AI Datasets: While trained on 1,500 construction awards, AI arbitrators may still lack the tacit knowledge, industry-specific “feel,” and evolving legal interpretation that seasoned human arbitrators bring.
5. Ethical and Accountability Gaps: Questions arise around accountability if an AI-drafted award proves flawed—particularly with regard to reasoning transparency, liability, and professional responsibility.
6. Over-reliance Risk: Without careful human oversight, AI recommendations could unconsciously influence arbitrators to defer to machine outputs, undermining independent decision-making.

AI in Arbitration: Governance and Ethics

As I discussed in my earlier article, https://www.kronicle.in/blog/arbitration-in-the-age-of-ai:-global-trends-and-indian-implications, the integration of AI into arbitration raises key issues: human oversight, transparency, and disclosure obligations. The AAA-ICDR’s March 2025 Guidance on Arbitrators’ Use of AI Tools underscores that AI should support, not replace, arbitrators’ judgment. Arbitrators must disclose AI use to maintain integrity and enforceability of awards.

Other institutions are also responding. The CIArb’s 2025 Guidelines stress that tribunals retain full responsibility for proceedings, while the SCC’s 2024 Guide emphasizes “effective human oversight.” Meanwhile, the ICC has launched a task force on AI, and HKIAC and SIAC are exploring partnerships and governance frameworks.

Enforceability of Awards by Indian Courts

In my personal view, an Indian court is unlikely to refuse enforcement of a foreign award solely because the tribunal used AI tools (e.g., an AI copilot that assists analysis and drafting) provided that:

• (a) a human arbitrator (or tribunal) exercised independent judgment and signed the award;
• (b) the composition/procedure complied with the parties’ agreement and the law of the seat; and
• (c) there was no breach of natural justice (notice and opportunity to be heard) or fundamental policy (e.g., secrecy/opacity that renders a party unable to present its case).

This flows from India’s proenforcement stance and the narrow reading of “public policy” for foreign awards under Section 48(2)(b) of the Arbitration and Conciliation Act, 1996 and the Supreme Court’s line of decisions in Renusagar, 1993 CaseBase(SC) 952, Shri Lal Mahal (2014) 2 SCC 433 and Vijay Karia [2020 CaseBase(SC) 1703].

However, refusal risks do arise if “AI arbitration” means that the AI itself is treated as the arbitrator (i.e., no human adjudicator), or if the AI’s use materially impairs due process (e.g., nondisclosure of AI reliance that prevents meaningful rebuttal), or if the tribunal composition/procedure deviates from the agreement or the seat’s law. Those issues map to Section 48(1)(b) (inability to present case), and Section 48(2)(b) (public policy).

The AAA-ICDR model is expressly human-validated (AI evaluates merits, drafts, human arbitrator reviews and signs). That design choice directly reduces Section 48 risks at the enforcement stage in India. However, a court could view machine adjudication without human judgment as offending the fundamental policy of adjudication (natural justice) or the most basic notions of justice. In Vijay Karia [2020 CaseBase(SC) 1703] the court framed “most basic notions of justice” narrowly, but opaque decision-making without human responsibility could qualify on extreme facts.

Treating an AI system as the arbitrator (with no genuine human adjudicator) risks refusal under Section 48(1)(d): composition not in accordance with the parties’ agreement or the law of the seat (most arbitration laws and rules presuppose natural persons as arbitrators who can make disclosures, deliberate, and sign awards).

What’s Next?

Initially focused on document-only construction disputes, the ICDR plans to expand AI arbitrator eligibility to other sectors and higher-value claims by 2026. Whether AI can replicate the nuanced expertise of human arbitrators in specialized sectors remains to be seen, but its role as an efficiency enabler is undeniable.

Over time, AAA leaders suggest expansion to other high-volume, lower-value disputes, with payer–provider insurance flagged as next. Longer term, the model could be generalized—but institutions will likely progress sector by sector, wherever datasets are rich and awards tend to be reasoned.

The AAA-ICDR’s AI arbitrator is not a robot judge and does not offend the fundamental policy of adjudication—it’s an augmented drafting and analysis engine embedded within a human-validated process. If it delivers the promised cost (30–50%) and time advantages while honoring due process, it may rebalance incentives toward resolution over compromise for a meaningful slice of cases. The institutions that combine quality data, rigorous oversight, and clear disclosure norms will set the pace for AI’s next chapter in arbitration. Sample Content