The United States proposed to China, during talks held in New York on Sunday, to create a mechanism that would allow the two countries to inform each other of incidents related to artificial intelligence that reach a level that affects national security, days before an expected meeting between US President Donald Trump and Chinese President Xi Jinping.
The proposal came three days after warnings issued by officials and experts during the Xiangshan Security Forum in Beijing about the risks that may accompany the increasing use of artificial intelligence in the military field, including reducing the time available for humans to make decisions with increasing reliance on more independent systems.
During the same period, American and Chinese security experts put forward proposed controls for the use of artificial intelligence militarily, warning of scenarios that might leave Washington or Beijing with a short time to determine whether an incident linked to sensitive military systems represents a deliberate attack by the other party.
These developments raise concerns about the repercussions of the increasing military use of artificial intelligence, especially with its ability to speed up the decision-making process. Will this increase the risk of miscalculation and push two countries into a confrontation that neither of them was seeking?
Faster decisions
Artificial intelligence has entered multiple areas of military action, from analyzing huge amounts of information and images to surveillance, monitoring threats, and supporting command and decision-making operations.
One of its most prominent military advantages lies in speed. Systems capable of processing massive amounts of data in a short period of time can give militaries a greater ability to detect and respond to threats, an advantage that is increasingly important on battlefields with a rapid pace of operations.
But the same advantage carries the opposite risk. The faster a threat can be detected and responded to, the less time humans have to verify information and assess an adversary's intentions before making a decision that may be difficult to reverse.
The American Brookings Institute indicates that the rapid entry of artificial intelligence into the military field, from electronic defense to weapons systems, information analysis, leadership and decision-making, opens new paths to errors or misuse that could lead to escalation.
The issue is even more sensitive when it comes to nuclear powers, where the cost of misinterpreting enemy activity can be much greater than on conventional battlefields.
When artificial intelligence makes mistakes
This scenario does not require the AI to independently make the decision to wage war. The problem may begin much earlier, when the information produced or analyzed becomes part of the series of decisions made by humans.
Systems that rely on artificial intelligence are not infallible, and an inaccurate classification of a target or incorrect analysis of information may present a misleading image to the decision maker, especially when he is required to determine the nature of the threat and respond to it quickly.
Another problem also arises in electronic operations, as it is sometimes difficult to determine who is responsible for the attack. It may come from a state, a non-governmental group, or an entity using artificial intelligence, while an operation may also be carried out in a way that suggests another state is behind it.
In such cases, the danger is not in the machine alone, but in the interaction between incomplete information and increased speed and pressure to make decisions. Responding to an attack that is incorrectly attributed to an opponent may lead to a series of measures and countermeasures that are difficult to stop later.
Risks affecting nuclear systems
Concern increases when artificial intelligence is linked to nuclear systems and electronic operations.
Last week, American and Chinese security experts warned that the interference of an artificial intelligence system in a nuclear command and control network, or its implementation of a military electronic operation, may leave the other party with a limited time to determine whether the rival government is behind what happened.
This does not mean that artificial intelligence has become capable of launching a nuclear war on its own. But its integration into military warning and analysis systems could add a new layer of uncertainty to crises for which the time available to manage is already limited.
The risks increase with the spread of more autonomous military systems, such as drones and unmanned ships. An unintended move or an automated response to an incident in an area witnessing military friction could be interpreted differently by the opposing party, which forces governments to contain the crisis after it occurs.
Keeping the decision in human hands
The risks of miscalculation prompted the search for controls to ensure the maintenance of human control in the most sensitive military areas.
The United States and China agreed in late 2024 that the decision to use nuclear weapons should remain under human control. American and Chinese researchers participating in a dialogue organized by the Brookings Institute and the Center for International Security and Strategy at Tsinghua University in China believe that this principle could constitute a starting point for broader controls.
Among the proposals are to keep the decision to carry out cyber attacks on nuclear command and control systems and sensitive infrastructure in the hands of humans, and to define a clear meaning for “human control” over military systems that use artificial intelligence.
Experts also suggest establishing a US-Chinese communication channel dedicated to incidents resulting from artificial intelligence errors or cyberattacks related to it, with the aim of preventing misunderstandings from turning into a broader crisis.
These proposals do not represent an agreement between the two governments, but they coincided with official movement in the same direction. US Treasury Secretary Scott Besent said after his talks with Chinese Deputy Prime Minister He Lifeng, on Sunday, that Washington has proposed a mechanism for reporting incidents related to artificial intelligence that reach a level that affects national security, and that Trump and Xi will discuss this mechanism during their meeting.
Race faster than the controls
But keeping humans in the decision-making chain does not completely solve the problem. The United States and China have not yet reached a common approach to regulating the military use of artificial intelligence, while their competition to develop more advanced capabilities continues.
Also, the fact that humans have the last word does not necessarily mean that they have enough time and information to verify the accuracy of what automated systems suggest. Increasing the speed of military systems may force people to make their decisions under greater pressure.
In addition to this, what Brookings researchers describe is a security dilemma resulting from the difficulty of knowing the true capabilities possessed by the opponent, or determining whether his military applications of artificial intelligence are defensive or offensive. Ambiguity may prompt each party to assume the worst scenario and develop corresponding capabilities.
Artificial intelligence does not eliminate the political calculations that lead countries to war or prevent them from war, but it changes an aspect of the deterrence equation between competing countries, by accelerating the detection and response to threats and reducing the time available to assess the situation before making a decision.
As the use of artificial intelligence increases militarily, it becomes increasingly important to maintain human control and give commanders sufficient time to verify information and correct errors before making a decision. The speed provided by these systems may be a military advantage, but in return it increases the importance of verifying the validity of the information before taking steps that might lead to unintended escalation.
During his upcoming meeting with his American counterpart, Monday, in New York, French President Emmanuel Macron intends to speak with Donald Trump “frankly” to try to “direct him towards the right decisions,” especially with regard to Ukraine, according to sources close to Macron.
The same sources added that Macron will try in this regard to persuade his American counterpart to “continue supporting Ukraine” and “increase pressure on Russia.”
The meeting is scheduled to be held at 17:35 local time (21:35 GMT), according to what Agence France-Presse reported.
In an interview with La Premiere TV station, Macron stated that he also intends to work to “alleviate tensions related to customs duties” with the United States.
The French President also expressed his desire to reach a “peaceful solution” to the Strait of Hormuz crisis, which is considered a “key” to calming tensions in fuel markets.
He confirmed his intention to raise the issue of war in Ukraine, because of its repercussions on the energy file and “the security of Europeans.”
The French President said that he seeks to consolidate the progress achieved during the G7 summit last June in the French city of Evian, as Donald Trump appeared to be approaching, even temporarily, European positions on this issue.
The Europeans demand that the United States maintain its commitment to settling the conflict, with American participation in the negotiations alongside Ukraine and Europe.
They also demand security guarantees, that is, an American commitment strong enough to deter any new Russian attack that might follow a potential ceasefire.
Trump's envoys, Jared Kushner and Steve Witkoff, went to Kiev for the first time in September to revive American mediation efforts in this conflict, which is considered the bloodiest in Europe since World War II.
They have previously visited Russia several times, noting that Russia has intensified its air strikes against Ukraine - especially the capital, Kiev - since the summer.
However, several rounds of tripartite negotiations have failed to achieve any tangible progress in the past.
Macron is heading to New York from Saint Pierre and Miquelon, a small French region in the North Atlantic Ocean, where the French President met on Sunday with Canadian Prime Minister Mark Carney. The two leaders defended the sovereignty of their countries and stressed the pursuit of a “deep” rapprochement in the face of “Trump’s intimidation.”
