Iu Kwan Yuen

Lawyers have spent the past few years learning an important lesson about generative artificial intelligence: never assume that what an AI system produces is correct. Authorities must be checked, quotations verified and professional judgment retained. A recent American decision identifies a different and potentially more difficult problem. What if the AI's output is unreliable not because the model hallucinated, but because someone deliberately manipulated the document fed into it?

Introduction In our article published in January 2026, “The Use of Artificial Intelligence in Arbitration”, we considered whether the Hong Kong legal framework could accommodate artificial intelligence (“AI”) in arbitral decision-making. At that time, the concern remained largely prospective. Subsequent developments now show that misuse of AI in preparing an award may give rise to a challenge to its validity or enforcement.