Sixteen arbitral institutions challenge EU AI Act high-risk classification guidelines

The classification of AI tools as high-risk under the EU AI Act imposes significant data governance, risk management, and human oversight requirements on the regulated entities. If the European Commission maintains its current broad definition of ADR bodies, arbitral institutions face increased administrative complexity and compliance costs that will likely be passed to parties in the form of higher arbitration fees. This regulatory tension reflects a broader challenge in the streaming and tech ecosystem: defining where automated assistance ends and legally binding decision-making begins. Industry observers should monitor the final Commission guidelines to see if administrative functions receive a formal exemption, which would shift the compliance burden directly to individual arbitrators and technology providers.

The joint letter from sixteen arbitral institutions arrives as the EU AI Act enters its critical implementation phase for high-risk systems. The International Chamber of Commerce, which operates the world's largest international arbitration body, has been among the most vocal institutions pushing back against broad AI classification. The ICC published guidance in early 2025 urging the European Commission to distinguish between administrative support functions and adjudicative decision-making in AI regulation, arguing that conflation of the two would impose disproportionate compliance burdens on dispute resolution bodies. The London Court of International Arbitration and the SCC Arbitration Institute, both signatories to the joint letter, have similarly raised concerns about how high-risk classification could affect their case management platforms.

The broader regulatory landscape for AI in legal and dispute resolution contexts remains unsettled across multiple jurisdictions. The Chartered Institute of Arbitrators has been developing its own framework for responsible AI use in arbitration proceedings, reflecting a growing recognition among professional bodies that self-regulation may be necessary if legislative frameworks prove too blunt. The American Arbitration Association and the International Centre for Settlement of Investment Disputes, both signatories to the joint letter, operate in jurisdictions where no equivalent AI legislation exists yet, creating potential compliance asymmetries for cross-border disputes involving EU-connected parties.

The technical question of where AI assistance ends and decision-making begins has parallels in other regulated industries grappling with the EU AI Act's scope. SAP has outlined a three-horizon framework for embedded business AI in complex manufacturing, distinguishing between automation of repetitive tasks, predictive optimization, and autonomous cross-company decision-making, a taxonomy that mirrors the arbitral institutions' argument that administrative AI tools should not be classified the same way as systems that influence substantive outcomes. The European Commission's final guidelines on high-risk classification are expected to determine whether arbitral institutions receive a formal carve-out or whether compliance obligations will cascade to individual arbitrators and their technology vendors.