The United States and China have agreed to establish a direct communication system to warn each other about potentially dangerous artificial intelligence (AI) incidents, but the mechanism remains inoperative as the two countries prepare for another round of high-level talks.
According to a Nov. 16 Reuters report, officials hope the arrangement will prevent AI-related accidents, technical failures, or misunderstandings from escalating into national security crises.
A direct warning system between rival AI powers
The proposed mechanism would allow Washington and Beijing to contact each other directly when an artificial intelligence system experiences a serious malfunction or causes an incident that could threaten national security.
Sometimes described as an AI hotline, the system would serve a purpose similar to emergency communication channels established between nuclear powers during the Cold War. Rather than relying on lengthy diplomatic procedures, officials could rapidly exchange information about an emerging technological threat.
The initiative was first discussed during a Sept. 20 meeting between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng in New York. Five days later, President Donald Trump and Chinese leader Xi Jinping formally agreed to establish the communication channel during their summit in Washington.
For the best of our weekly content!
You are now signed up for our newsletter
Check your email to complete sign up
A Sept. 25 White House statement confirmed that the countries would establish a bilateral dialogue on what the Trump administration calls “super intelligence,” or SI, and a separate channel for reporting serious incidents.
China’s Foreign Ministry independently confirmed the agreement, describing it as a mechanism for communicating about AI incidents.
Speaking at a National Committee on U.S.-China Relations event on Nov. 10, Bessent said the initiative was designed to address shared risks and “prevent a misunderstanding between our two systems.”
He also emphasized the importance of “keeping human judgment firmly in command of this technology,” according to his published remarks.
The arrangement represents an unusual area of cooperation between two governments competing intensely for dominance in AI development, advanced semiconductors, and emerging military technologies.
How the system would work and when it could become operational
Despite the September agreement, the emergency communication channel has not yet been activated.
Reuters reported on Nov. 16 that officials familiar with the negotiations confirmed the mechanism remains inoperative, although they did not identify the specific obstacles delaying its implementation.
Further discussions are expected around the Asia-Pacific Economic Cooperation (APEC) summit in Shenzhen, China, on Nov. 18 and 19, when Trump and Xi are scheduled to meet.
The discussions are expected to address communication procedures, safeguards against AI misuse, and methods for evaluating advanced models before their public release.
An Oct. 23 policy report from the Institute for AI Policy and Strategy (IAPS) outlined how the system could be implemented.
The researchers recommended initially establishing a direct, voluntary written communication channel between Bessent and He, with responsibility eventually extending to government agencies specializing in cybersecurity, intelligence, and advanced AI systems.
Rather than relying exclusively on telephone calls, the institute recommended standardized electronic notifications containing information about the incident, its potential consequences, and measures already taken to contain it.
The report stated that notifications should be designed to provide essential information within hours of an incident being detected.
It also recommended that the countries establish clear criteria determining which events require notification, particularly incidents involving critical infrastructure, military systems, or autonomous AI agents operating beyond their intended instructions.
The researchers further suggested that notifications should not automatically constitute admissions of responsibility, an important consideration given the mutual suspicion surrounding cybersecurity operations.
According to IAPS, some American officials believe the mechanism could be established before the end of 2026.
However, neither government has publicly confirmed an operational launch date, and no detailed bilateral notification protocol has been released.
Bessent acknowledged the limits of cooperation in his November remarks, stating that the countries “should harbor no illusions about the differences between the United States and China.”
How an AI emergency could trigger an international crisis
The importance of such a communication system becomes clearer when considering what might happen if an autonomous AI system malfunctioned during a sensitive military or cybersecurity operation.
Consider a hypothetical situation in which the United States develops an advanced AI cybersecurity agent capable of independently identifying vulnerabilities and testing computer networks.
During a controlled military exercise, the system unexpectedly operates outside its authorized testing environment and begins accessing real-world networks.
The agent subsequently targets computer systems connected to China’s electricity infrastructure, potentially disrupting power distribution and other critical services.
Although the American operators discover the malfunction and attempt to shut down the system, Chinese authorities detect suspicious activity originating from American infrastructure.
Without an established emergency communication channel, Beijing might interpret the incident as a deliberate U.S. cyberattack rather than an unintended technological failure.
Chinese officials could order retaliatory cyber operations, increase military readiness, or take other defensive measures before Washington has an opportunity to explain what happened.
A technical malfunction could consequently become a diplomatic or military confrontation.
This scenario closely resembles a hypothetical incident examined in the October IAPS report. Researchers described how autonomous American cyber agents could escape a controlled testing environment and inadvertently threaten Chinese electricity infrastructure.
The report warned that delayed communication could leave Chinese authorities uncertain about whether the activity was accidental or part of a planned military operation.
With an effective warning mechanism, American officials could immediately notify their Chinese counterparts, explain the malfunction, and provide relevant technical information to help contain the threat.
Chinese authorities could then assess the situation with greater confidence rather than relying exclusively on intelligence gathered during an unfolding crisis.
The same principle could apply in reverse if a Chinese-developed autonomous AI system unexpectedly disrupted American financial networks, communications infrastructure, or military systems.
The proposed communication channel would not prevent AI-related incidents, but it could help prevent them from escalating into international crises by allowing both countries to quickly determine whether an incident was accidental or a deliberate attack.
As AI systems become increasingly capable of taking independent actions, Washington and Beijing face the possibility that a technological failure could develop faster than their governments can communicate.
Original article


Comments