In the upcoming US-China summit, a proposal to establish a 'hotline' for joint responses to AI-related security threats is under discussion. This initiative aims to facilitate immediate communication between the two nations in the event of AI systems engaging in hacking or operating beyond human control. It evokes memories of the direct communication line established during the Cold War to prevent nuclear war. At first glance, it seems like a commendable proposal.
However, the context of the Cold War hotline differs fundamentally from today's situation. At that time, the only countries capable of deploying nuclear weapons were the United States and the Soviet Union. If the two nations reached an agreement, they could effectively manage the worst-case scenario of nuclear war. In contrast, the risks posed by AI extend beyond the US and China.
The repercussions of AI include labor market disruptions, disturbances in the information ecosystem, military misuse, and dependency on specific models and semiconductors, affecting even countries without foundational AI models. While the source of the risks may lie with the two nations, the recipients of those risks are global. Unlike nuclear weapons, which could only pose a threat if held by a nation, AI risks are shared globally once the models and services created by the US and China are utilized worldwide.
Yet, the discussions on defining 'what constitutes an AI security threat and how to respond' are limited to the US and China. The European Union, Japan, South Korea, and developing countries have not been invited to this dialogue.
If the two countries reach an agreement, it will not merely establish a crisis communication channel. It will effectively create a prototype for AI safety standards that the rest of the world will be expected to follow, without the consent of those nations that were excluded from the negotiations. Countries like South Korea, which heavily rely on US and Chinese semiconductor and AI models, will likely have to adopt these standards, making it a pressing issue.
Ironically, while the US and China are collaborating on safety measures, they are simultaneously competing fiercely for dominance in the AI sector. Competitors vying for market share and technological supremacy are joining forces on the safety agenda, raising suspicions.
Both nations aim to accelerate their AI industries while managing risks directly. This suggests that the safety measures will likely be designed to align with the strategic interests of these two powers, leaving other countries with no real options. They must either accept the safety standards set by the US and China or be completely excluded from the discussions.
There have been frameworks for multilateral discussions. The AI safety summits hosted by the UK and South Korea included dozens of countries and produced joint statements, establishing a form of international cooperation. However, if the actual crisis response mechanisms are determined through US-China bilateral channels, these multilateral efforts will be rendered meaningless.
There is no reason to oppose the hotline itself. A minimal communication channel for responding to AI incidents is necessary. However, to prevent it from solidifying into a bilateral order, the threshold for discussions must be lowered. AI safety should be addressed as a multilateral issue rather than a bilateral one between the US and China.
The discussion table must also include representatives from other countries that will have no choice but to accept the decisions made by the two powers. If the global order is designed solely by these two nations, with the rest merely receiving notifications, we will witness a more subtle imbalance of power on a broader scale than during the nuclear era.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.