China and the United States plan to establish a reporting mechanism for high-risk artificial intelligence incidents to prevent major national security risks

Abstract:

Although the United States and China have strategic competition in many fields, according to technology media reports, the United States is seeking to establish a security notification mechanism with China to issue early warnings to each other in the event of dangerous artificial intelligence incidents that may pose serious national security threats.

This idea was jointly proposed by the Atlantic Council and Security Exploration International. The core goal of this mechanism is to ensure that the two countries can contact each other in a timely manner before serious artificial intelligence security incidents evolve into geopolitical crises or military conflicts by establishing special early warning channels. The report pointed out that due to the rapid evolution of artificial intelligence technology, automated systems are very likely to cause unexpected and serious consequences without direct human control, so establishing such communication bridges is crucial.

According to the proposal, the notification mechanism focuses on four major categories of high-risk scenarios:

First, the autonomous AI system loses control and causes serious real-world damage;

Second, AI safety accidents involving weapons of mass destruction such as nuclear weapons, chemical weapons or biological weapons;

Third, the accidental leakage of high-risk AI models or related core technologies;

Fourth, large-scale AI network attacks target critical infrastructure and cause devastating blows.

Expert analysis pointed out that the idea of ​​establishing this mechanism is very similar to the "red line" direct phone call established by the United States and the Soviet Union during the Cold War to prevent the outbreak of nuclear war. Although technology companies and research institutions in China and the United States have already held preliminary dialogue on AI security standards and risk management, if this cooperation can be formally elevated to a government-level security mechanism, it will undoubtedly significantly reduce the risk of major international crises caused by technical misjudgments and provide important support for global AI governance.