The DIS has joined 15 leading arbitral institutions in submitting a joint response to the European Commission's public consultation on the draft Guidelines for high-risk AI systems under the EU AI Act.
In their submission, the signatories call on the Commission to clarify that arbitral institutions acting solely in an administrative capacity should not be classified as deployers of high-risk AI systems. The arbitral institutions advocate that they “administer proceedings thereby rendering administrative and organisational services to the parties and to the arbitral tribunal”. It is pointed out that they do not decide the merits of the dispute, notwithstanding that an arbitral institution, or one of its constituent bodies, may use terms such as “court”, “commission”, or similar designations in its name.
The Commission’s Draft Guidelines on the classification of high-risk AI systems are intended to help providers, deployers and market surveillance authorities determine whether an AI system qualifies as high-risk under Article 6 of the EU AI Act. The Guidelines are expressly non-binding and represent the European Commission's interpretation of the AI Act pending any authoritative interpretation by the Court of Justice of the European Union.
The joint submission forms part of the European Commission's ongoing consultation process. The final Guidelines are expected to be adopted by the end of 2026.
The letter welcomes the European Commission’s efforts to provide guidance on the classification of high-risk AI Systems under the EU AI Act. However, the institutions recommend targeted refinements to the Draft Guidelines to ensure that regulatory obligations are assigned to the appropriate actors within international arbitration.
The letter’s principal argument is that arbitral institutions should not be treated as “ADR bodies” for the purposes of the high-risk AI provisions where they merely administer arbitral proceedings. Unlike arbitral tribunals, institutions do no decide disputes on their merits. Rather, they provide administrative and organisational services such as case management, tribunal appointment support and financial administration. Consequently, classifying arbitral institutions as deployers of high-risk AI systems would place legal obligations on the wrong actor.
The signatories further point out that the Draft Guidelines already distinguish between judicial administration and judicial decision-making, excluding administrative functions from the scope of the AI Act. They argue that the same reasoning should apply to arbitral institutions, whose role is similarly limited to supporting, rather than adjudicating, proceedings. This interpretation is also consistent with Recital 61 of the AI Act, which excludes AI systems used solely for ancillary administrative activities.
In addition, the letter recommends revising the description of investment dispute bodies in the Draft Guidelines. It notes that investment arbitration typically involves disputes between investors and states, rather than disputes “between businesses”, making the current wording inaccurate.
The signatories therefore invite the Commission to clarify that, where Annex III(8)(a) of the AI Act applies, the relevant obligations should attach to AI systems used by or on behalf of arbitral tribunals in their adjudicative functions – not to AI systems used by arbitral institutions for administrative case management. They also encourage the Commission to reconsider the characterisation of investment dispute bodies before finalising the Guidelines.
Carla Mackensen