Who Owns the Risk When AI Gets It Wrong? Confidentiality, Governance and Human Judgment at LIDW 2026

The second half of the LIDW 2026 discussion addressed the governance challenges of artificial intelligence in arbitration, moving beyond technical limitations to complex questions of responsibility. A key debate centered on confidentiality and legal professional privilege, with Mr. Charlie Morgan warning that candid interactions with AI systems create new, discoverable records that may lead to disputes. Mr. Dmitri Evseev clarified that the critical issue is often data retention policies rather than model retraining, and he argued that locally run models do not risk privilege. The panel also examined institutional responses, noting a lack of engagement from arbitral bodies due to the rapid pace of technological change, though the ICC AI Task Force is expected to publish recommendations. Regarding AI-assisted arbitrators, Mr. Morgan expressed skepticism about the market's readiness for such a shift. The discussion then turned to commercial realities, where Mr. Stéphane Altounian highlighted a billing paradox: while clients expect cheaper services due to AI, firms are investing heavily in technology, and the volume of legal work is actually increasing. The session concluded with a consensus that AI's value lies in enhancing access to information through retrieval and verification, rather than replacing human judgment. The panellists emphasized that successful use of AI requires active engagement, grounding, and context management, ultimately framing the challenge as one of technological competence rather than displacement.

(Source：Scconline)