Human Dignity, Frontier AI, and the Limits of Experimentation

AI has been said to potentially pose an existential risk to humanity. While the human rights community focuses on the immediate harms to human rights, this article argues that frontier AI, namely highly capable AI models that can exhibit a general mastery of tasks across different domains, can raise a larger existential risk concern to human rights. Bearing in mind the experimental nature of frontier AI and the eventual aim of designing artificial general intelligence, this paper addresses the question of whether human dignity sets a hard limit as to what can be experimented upon individuals and society. It examines lessons from biotechnology regulation and social media experimentation which challenged related conceptions of human dignity in order to unpack and answer the question. It finds that respect for human dignity necessarily entails that procedural hard limits be set as safeguards for proceeding with the experimental nature of such technology and examines the different modalities available.