Towards a Universal Convention on Artificial Intelligence for Humanity
摘要
The global governance of Artificial Intelligence (AI) is fragmented, with mostly non-binding principles, codes of conduct, guidelines, and declarations predominating at the national or regional levels. Diverse national interests, a lack of international consensus, and the changing pace of advancements in AI technology drive this fragmentation. The desire for regulatory sovereignty, inadequate international frameworks, geopolitical tensions, and multiple stakeholders exacerbate the issue. While existing initiatives share the common goals of promoting trustworthy AI, ensuring safety and security, respecting privacy, preventing bias and discrimination, promoting accountability and transparency, and encouraging human oversight, their non-binding nature, uneven applicability, and lack of coherence render them insufficient to ensure AI benefits all humanity, particularly underrepresented groups in the Global South. Without a unified, binding framework, the current situation risks exacerbating inequalities and leaving vulnerable populations without sufficient protections or benefits from AI advancements. This chapter aims to initiate meaningful dialogue and action towards a unified global approach to AI governance through a Universal Convention on Artificial Intelligence for Humanity (UCAIH). Given the uneven technological advancements and varied value systems among nations, the chapter will highlight the inherent difficulties and contradictions in establishing a universal AI convention but argue that this is an opportune time to conceptualize such a framework. The chapter highlights the broad contours of the proposed convention, advocating for the inclusion of provisions on ethical governance, transparency, privacy, security, and equitable distribution of AI benefits. It calls for a specialized United Nations (UN) agency to oversee the development and implementation of this convention, leveraging its global reach and expertise. Only a binding convention can ensure international collaboration, addressing risks associated with uneven technological advancements and divergent regulatory approaches.