Singapore proposes UN framework for AI safety rules

Singapore has proposed a United Nations framework convention on artificial intelligence safeguards, as governments face risks such as autonomous systems going out of control and the potential development of weapons of mass destruction through AI. Foreign Affairs Minister Vivian Balakrishnan stated that the UN should play a central role in establishing common AI rules due to its universal membership.

Dr. Balakrishnan outlined three main risk categories: losing control of autonomous systems, rogue actors using AI for weapons, and economic, social, and political upheaval. He emphasized the need for rigorous testing, limits on autonomous systems, and human intervention when systems behave unexpectedly. Singapore's approach involves testing AI systems before deployment, restricting their capabilities, and ensuring human accountability.

The minister stressed that a global approach is necessary due to the cross-border nature of AI risks, but warned that trust among nations is lacking in the current strategic competition landscape. He suggested greater scientific cooperation, shared understanding of AI capabilities and risks, comparable testing methods, and rapid reporting of AI incidents across borders.

Dr. Balakrishnan also called for multilateralism in a fragmented world, urging adherence to existing international law and reform of global institutions like the UN. He highlighted Singapore's interest in the promise and perils of AI and expressed cautious optimism about the ongoing discussions on AI safety.