WAIC 2026 Roundtable: Decision-Making Involving Life and Death Should Never Be Delegate to AI, Humans Can Authorize Actions But Not Responsibility

According to Dynamics Beating monitoring, Xue Lan, Dean of the Institute for Artificial Intelligence Governance at Tsinghua University, Nicholas B. Dirks, President and CEO of the New York Academy of Sciences, and Mark Nitzberg, Executive Director of the Human-Compatible AI Center at the UC Berkeley AI Research Lab, participated in a roundtable discussion at the 2026 World Artificial Intelligence Conference, focusing on AI agent governance. The guests pointed out that AI is transitioning from "assisted judgment" to "autonomous action," becoming an intelligent agent that acts on behalf of humans, presenting a new manifestation of the traditional principal-agent problem in the digital age. Unlike human agents, AI faces alignment issues and, as a black box, cannot bear legal responsibility; therefore, accountability must shift to the entire execution chain, including developers, deployers, regulators, etc.

The guests reached a high degree of consensus: decisions involving life-and-death consequences, scenarios where mistakes are irreversible, and all issues involving ethics and value judgments must never be entrusted to AI. Humans can authorize AI to act, but cannot authorize AI to be responsible. Each authorization should be revocable, each action should be traceable, and the speed of relinquishing agency should never exceed the speed at which humans verify AI capabilities.

On the security front, the guests proposed that building trustworthy AI should meet three main engineering characteristics: a solid foundation, operational transparency, and controllability during use. In terms of institutional development, a globally unified AI security assessment standard, a mutually recognized testing system, and an incident data-sharing mechanism need to be established to define the red lines of AI development, set up early warning monitoring mechanisms, and promote AI security as a global public good. The boundary of AI governance is not the endpoint of technology but the starting point for humans to redefine their own values, responsibilities, and the direction of civilization.

---------------------------------

Click on the original text link below to join Dynamics Beating's Feishu AI News Channel, which continuously monitors global AI hotspots and news 24/7.