UNCITRAL Model Law on Automated Contracting and the Road to Automated and Autonomous Arbitration

As global commerce rapidly embraces automation via frontier technologies such as blockchain and artificial intelligence (“AI”), international arbitration must evolve accordingly. Although international efforts, such as UNCITRAL’s stocktaking on dispute resolution in the digital economy, mark important progress, arbitration law still lags behind. While these laws adapt, we can look to other legal instruments that have already taken the leap, particularly the UNCITRAL Model Law on Automated Contracting (MLAC).

This post examines the MLAC to determine if it could provide a legal foothold for automation in arbitration and forms part of a broader study into the MLAC's potential implications for arbitration.

Model Law on Automated Contracting

Adopted in July 2024, the MLAC establishes a legal framework for recognizing automated actions in the formation and performance of contracts. Its core principle is non-discrimination against automation, an embedded rule applying to all contracts. This is reflected in Article 5, which provides that a contract is not denied validity or enforceability solely because an action towards its formation or execution was performed through an automated system, without human intervention. At the same time, the provision preserves party autonomy by allowing parties to expressly limit or exclude such automated actions.

MLAC and Arbitration as Contractual Performance

The relevance of the MLAC to arbitration stems from the nature of the arbitration agreement. As the arbitration agreement is itself a contract, and the arbitral process constitutes its performance, the MLAC’s provisions could, in principle, extend to arbitration. When read together with Article 5, this suggests that the MLAC could provide a legal basis for the automation of specific procedural actions in arbitration.

Arbitration, however, has a sui generis nature. While the arbitration agreement is recognized as a distinct contract under the doctrine of separability, it also carries a jurisdictional dimension. The tribunal’s authority derives not only from party consent but also from the applicable arbitration law.

The contractual aspect of the arbitration agreement is embodied in procedural actions that give effect to party consent, such as the giving of notice, the constitution of the tribunal, exchanges of communications, and other administrative steps undertaken by the parties or, where authorized, by third parties such as arbitral institutions. These actions constitute performance of the parties' contractual obligations to initiate and cooperate in the arbitral process.

The jurisdictional aspect is engaged once the tribunal is constituted. From that point onward, the tribunal exercises adjudicatory authority conferred by the applicable arbitration law, rather than by the contract alone.

Therefore, when considering the legitimacy of automated systems in arbitration, the MLAC is most relevant to the contractual performance layer of the arbitration agreement. The precise extent of the MLAC's application in this context remains subject to deeper research, including consideration of competing theoretical approaches to the nature of arbitration (contractual, jurisdictional, etc.).​

Procedural Automation, Delegation, and Attribution in Arbitration

Automated systems are physical tools or software programs that follow and facilitate pre-defined rules, without requiring human intervention. For example, a software program may be designed to finalize and notify hearing dates to parties and arbitrators, by identifying available dates, finalizing the schedule, and sending notifications via email, all without human review. In doing so, it follows a set of pre-programmed instructions or algorithms, with one such rule being the selection of only those dates that are available in common to all persons involved. As selection of hearing dates forms part of procedural performance of an arbitration, such automation could fall within the MLAC’s scope.

There are several automation tools that are now being relied on in arbitration, such as OPUS 2 and DIS eFile. Many such systems are operated not by the parties themselves, but by arbitral institutions or tribunals. Because procedural functions may be delegated through the arbitration agreement and incorporated institutional rules, the use of automated systems by an institution or tribunal can be characterized as part of the contractual performance of the arbitration agreement.

When read alongside the MLAC’s non-discriminatory principle, a procedural act should not be denied validity solely because an automated tool was operated by an institution or tribunal, unless the parties have expressly opted out. Although not mandatory, legal certainty may be enhanced by the parties expressly acknowledging the use of automation in arbitration, either in the arbitration agreement itself or by incorporating rules that facilitate such use.

In this regard, it is recommended that arbitral institutions and organizations consider adopting dedicated rules on the use of automated systems, including provisions on attribution, which parties could then incorporate by reference into their arbitration agreements. This proactive approach aligns with Article 7(1) of the MLAC, which enables parties to establish a tailored attribution framework and avoid default attribution under Article 7(2), pursuant to which attribution falls on the user of the automated system.

In the arbitration context, the user could be an institution, tribunal or any third party operating the system. Importantly, the automated system itself cannot be the bearer of attribution. Instead, actions or outputs generated by the system must be linked back to the human who operates or controls it. This view respects the principle that procedural acts must be attributable to a responsible legal person or entity. This analysis considers attribution solely under the MLAC; applicable contract law, arbitration law, or other regimes may lead to different attribution outcomes.

While the MLAC provides a foundation for procedural automation, arbitration laws remain the primary legal framework governing the arbitral process, with party autonomy operating within, and not above it. Where mandatory provisions restrict automation, those provisions will prevail. For instance, mandatory provisions on due process and equal treatment reflected in Articles 18 and 17 respectively of the UNCITRAL Model Law on International Commercial Arbitration must prevail. These requirements ensure procedural fairness, which is essential both for the award’s enforceability under the New York Convention and for the arbitration's legitimacy as a credible alternative to courts. The MLAC complements this framework by limiting its own scope, as illustrated in Article 7(4), which provides that nothing in the attribution framework affects the application of any rule of law that may govern the legal consequences of attributing an action carried out by an automated system to a person.

From Automation to Autonomous Arbitration

Having established how the MLAC can support procedural automation in arbitration, the next question is how such processes could evolve toward greater arbitral autonomy. In this context, the term “autonomy” is used in relation to “autonomous arbitration”, where arbitration can operate free from state intervention, as discussed here. Frontier technologies such as smart contracts, oracles, and AI, which fall well within the scope of the MLAC, could play a role in facilitating technological automation, which can in turn progress toward achieving the goal of autonomous arbitration.

For instance, embedding an arbitral award into a smart contract could enable automated enforcement of the decision, reducing the need for state intervention. While fully autonomous arbitration is conceptually appealing, it demands a careful balance between technological optimization and the legitimacy conferred by law and policy.

As highlighted above, mere application of the MLAC does not automatically legitimize all elements of automation in arbitration. Fully automated decision-making, such as arbitration awards rendered entirely by machines, are beyond its scope. The MLAC only validates automated actions in contract formation or performance, as reflected in its deliberate avoidance of the term “decision” in Article 5.

Nonetheless, progress is being made toward AI-assisted decision-making in arbitration.  For instance, the recently released  AI Arbitrator by AAA-ICDR is designed to arbitrate document-only construction cases where efficiency and speed are critical. A human arbitrator will review the AI’s reasoning, evaluate and, if needed, revise the outcomes before the decision is finalized, demonstrating the continued importance of human oversight even as automation expands. 

Conclusion

Arbitration is a nuanced process governed by specialized laws, carefully designed to safeguard procedural integrity and equal treatment. This very precision and emphasis on legitimacy, however, often means that arbitration laws are slower to adapt to technological change. As arbitration-specific laws gradually adapt, it may be necessary to look beyond them to find the legal validation and confidence needed to support innovation.

Historically, arbitration has incorporated technical processes before formal legal recognition. Electronic communications and contracting for example, were widely used long before the 2006 amendments to the UNCITRAL Model Law on International Commercial Arbitration formally recognized data messages for the “in writing” requirement. Earlier instruments such as the UNCITRAL Model Law on Electronic Commerce (MLEC, 1996) and the United Nations Convention on the Use of Electronic Communications in International Contracts (2005) laid the groundwork by establishing the legal equivalence of electronic and written form, which was later reflected in arbitration law.

Similarly, the MLAC’s non-discrimination principle provides a timely legal foothold for incorporating automation into arbitration. Grounded in the contractual nature of the arbitration agreement, it supports automated procedural steps such as scheduling or enforcement without requiring express prior consent.

Since the MLAC is a Model Law awaiting state adoption, now is an ideal moment to assess its implications for arbitration and to contribute to ongoing deliberations, especially in the context of the work of UNCITRAL Working Group II. This post seeks to prompt that dialogue and to map both the possibilities and limits of the MLAC in guiding arbitration’s evolution. The significance of this validation may extend beyond procedural efficiency, as it could also pave the way toward a legitimized form of autonomous arbitration.