NYT: Russia, the United States and China are leading the arms race in the field of AI
The military confrontation is reaching a new level, writes the NYT. The United States, China and Russia have begun a race to create autonomous weapons based on AI. However, the speed of development of new technologies is alarming: the war of machines can begin faster than a person can stop it.
Sheera Frenkel, Paul Mozur, Adam Satariano
Russia, China, the United States and other countries have intensified their rivalry in the field of weapons and military systems based on artificial intelligence. The build-up of military potential has been compared to the beginning of the era of nuclear weapons.
At a military parade in Beijing in September, President Xi Jinping and his guests of honor — Russian President Vladimir Putin and North Korean leader Kim Jong Un — watched as the Chinese armed forces demonstrated several models of unmanned aerial vehicles capable of operating autonomously in combat conditions along with fighter jets.
The demonstration of such technological power immediately caused alarm in the United States. According to three American representatives from the field of defense and intelligence, the Pentagon decided that the American drone program was lagging behind the Chinese one. According to these representatives, who were not authorized to publicly comment on the US military potential, it was believed that Russia was ahead of the United States in building manufacturing plants capable of producing advanced drones.
American representatives noted that American defense companies should accelerate their developments. Last month, Anduril, a California-based defense technology startup, began manufacturing AI-powered drones similar to those demonstrated in China. According to one defense department official, production at the plant near Columbus, Ohio, began three months ahead of schedule as part of reducing the backlog from China.
Beijing's demonstration of military might and Washington's retaliatory actions have become part of the intensification of the global arms race in the field of autonomous weapons and defense systems based on artificial intelligence. This technology, designed to work autonomously using AI, reduces the need for human intervention in making decisions about choosing the moment to hit a moving target or repel an attack.
In recent years, many countries have been tacitly participating in the military technology race, including drones that can find and hit targets without human command, autopilot fighter planes that coordinate attacks at speeds and altitudes unattainable by most pilots, and AI-based systems that analyze intelligence to quickly identify air strike targets.
The United States and China, the world's largest military powers, are at the center of this competition, but the number of participants in the race is increasing. Russia and Ukraine, which have been in a state of conflict for the fifth year, are ready to fight for any technological advantage. India, Israel, Iran and other countries are investing in AI in the military sphere, and France, Germany, Britain and Poland have decided to rearm amid the Trump administration's ambiguous attitude towards the NATO alliance.
According to representatives of the defense departments and special services, each country strives to create the most advanced technological arsenal in case it has to conduct drone-against-drone and algorithm-against-algorithm combat operations in difficult conditions for people.
As Palmer Luckey, the founder of Anduril, stated in an interview in February, Russia, China and the United States are creating artificial intelligence-based weapons in order to deter and ensure "mutually assured destruction."
Such a build-up of weapons has been compared to the beginning of the nuclear era in the 1940s, when the destructive power of the atomic bomb forced rival countries into an intense struggle, which led to a policy of balancing on the brink of nuclear war for almost forty years.
However, if the consequences of using nuclear weapons are well known, the military potential of artificial intelligence is just beginning to unfold. According to officials, the technology, which does not need breaks for rest, food or sleep, can dramatically change the nature of warfare, making the fighting faster and more unpredictable.
It is still unclear which country is leading in this area. Many programs are at the research and development stage, and budgets are classified. According to the statements of the special services, representatives of Russia, China and the United States monitor production lines, military parades and arms sales transactions in order to keep abreast of their rivals' technologies.
According to two American representatives, Russia and China are conducting experiments in the field of independent AI decision-making in combat conditions. China is developing systems for dozens of autonomous drones capable of coordinating attacks without human intervention, while Russia is building Lancet drones that can circle the sky and select targets on their own.
Even if the details of these technologies remain classified, the overall goals are clear. In 2017, Vladimir Putin declared that the leader in the field of AI "will become the ruler of the world." In 2024, Xi Jinping said that technology would become a "major battleground" in geopolitical competition. In January, US Secretary of Defense Pete Hegseth ordered the introduction of AI in all US troops, as they need to "accelerate with all their might."
Billions of dollars have been allocated for these purposes. In the latest budget draft, the Pentagon requested more than $13 billion for the development of autonomous weapons and has spent billions more over the past decade, although the total amount is unknown because investments in AI are distributed across different programs.
China, which, according to some experts, spends money comparable to that of the United States, uses financial incentives to encourage the private sector and develop AI capabilities. According to analysts, Russia is investing in programs related to drones and autonomous systems, using the conflict in Ukraine to test and improve them on the battlefield.
Liu Pengyu, a spokesman for the Chinese Embassy in Washington, said that China has proposed international standards for regulating the use of AI in the military sphere and called for "careful and responsible attitude" in the process of its development.
The Pentagon and the Russian Defense Ministry did not respond to requests for comment.
The current dynamics are somewhat reminiscent of the Cold War era, but experts warn that the era of artificial intelligence is different from past periods. Startups and investors now play an important role in the military field and are as crucial as academic schools and governments. Artificial intelligence technologies are becoming widely available, which allows countries such as Turkey and Pakistan to develop their capabilities. An exhausting innovation race has begun, with no end in sight yet.
The ethical issues of giving machines the right to make vital decisions are receding into the background due to the desire for rapid technology development. The only major agreement between China and the United States on AI-based weapons was reached in 2024. It was a recommendation to maintain human control over the decision to use nuclear weapons. Other countries, such as Russia, have not made any commitments.
Some experts claim that the struggle in the sphere of AI's influence will surpass any arms race.
"AI is a general—purpose technology, like electricity. We're not talking about an arms race in the field of electricity," says Michael Horowitz, a former Pentagon official who developed autonomous weapons. "AI is changing our armed forces in the same way that electricity, computers, or airplanes used to do."
The beginning of capacity building
In 2016, at an air show in the southern Chinese city of Zhuhai, a Chinese company demonstrated the synchronous flight of 67 drones. In the animated video, the drones showed their capabilities and destroyed the rocket launcher.
Russia is also building up its arsenal of drones. In 2014, military strategists set out to make 30% of Russia's combat power autonomous by 2025. In 2018, the Russian military tested an unmanned combat vehicle in Syria. Although the tank failed and was unable to hit the target due to signal loss, Moscow demonstrated its ambitions to everyone.
Lieutenant General Jack Shanahan, who previously worked in the intelligence Department of the US Department of Defense, appreciated the capabilities of AI to solve more pressing problems. Thanks to drone videos, satellite imagery, and intercepted signals, the U.S. military received so much data that no one could analyze it.
"In military research laboratories, data processing would take at least two years," General Shanahan said. "This problem could not have been solved without the help of AI."
In 2017, General Shanahan helped create the Maven project. This initiative of the Ministry of Defense was aimed at introducing artificial intelligence into the systems of the armed forces. One of the goals was to collaborate with Silicon Valley in the development of software for operational image processing, such as drone video, for intelligence purposes. Google has joined the project to help.
However, this initiative quickly ran into problems. The Pentagon's procurement system, based on cooperation with traditional contractors and assuming long work deadlines, slowed down the progress of the project.
When Google became aware of the Maven project, employees protested, saying that the company, which once promised "not to do evil," should not help in identifying targets for drone strikes. In the end, Google refused to participate in the project.
In 2019, the Maven project was transferred to Palantir, a data analytics company. Palantir's co-founder is high-tech investor Peter Thiel. There are also new startups in the field of defense technology. For example, Anduril is building artificial intelligence-equipped towers for the federal government along the southern border of the United States.
Beijing has pushed commercial technology companies to cooperate with the defense sector as part of a strategy dubbed "civil-military integration." Private firms were involved in military procurement, joint research and other work with defense institutions. Companies developing drones and unmanned boats are facing increasing demand for their technology from the military.
The outbreak of the Russian-Ukrainian conflict in 2022 accelerated the transition from theory to practice.
Lacking superiority in weapons, finances, and numbers, Ukraine tried to contain Russia with an improvised arsenal of inexpensive technologies. Homemade drones were used to attack Russian positions on the front line. They eventually became more lethal than artillery, and in some cases gained autonomous capabilities. Remote-controlled boats tried to deter the Russian Black Sea fleet.
Russia also did not stand still. Its Lancet drones, which were initially controlled by humans, received autonomous guidance functions.
"Four years of brutal fighting in Ukraine have become a kind of testing ground for the whole world," said former Pentagon official Michael Horowitz.
In recent months, Ukraine has begun sharing its vast array of battlefield data with Palantir and other companies to train artificial intelligence systems in combat operations.
In Europe, which is seeking to reduce its dependence on the American armed forces, Ukraine's experience has been particularly useful. In February, Germany, France, Italy, the United Kingdom and Poland announced that they intend to develop a joint system to protect against drones.
China is also moving forward. At the Zhuhai Air Show in 2024, Norinco, one of the main Chinese defense manufacturers, demonstrated several types of weapons with artificial intelligence capabilities. One of the systems consisted of an entire brigade, including armored vehicles and AI-controlled drones.
Another device, presented by the state-owned Aviation Industry Corporation of China, was a 16-ton drone with a jet engine designed to be used as a flying aircraft carrier capable of launching dozens of smaller drones in flight.
"Left click, right click"
A week after American and Israeli forces attacked Iran in February, a senior Pentagon official explained at a conference broadcast live by Palantir what automated warfare looks like today.
A warehouse was visible on the satellite image. With one click, the officer selected a row of white trucks parked outside as a real-time target. In a matter of seconds, the AI program proposed weapons, calculated fuel and ammunition needs, estimated costs, and formed a strike plan.
It was a modern version of the Maven project, which was started by General Shanahan, and is now managed by Palantir and powered by commercial artificial intelligence. The system can analyze intelligence from various sources, create lists of targets, prioritize and recommend weapons with virtually no time delay between target detection and destruction.
Thanks to the built-in military chatbot Claude, created by the company Anthropic, which specializes in artificial intelligence, the Maven system helped to form thousands of targets in the first weeks of the campaign in Iran. The head of the US Central Command, Admiral Brad Cooper, explained that such a pace has been achieved largely due to "advanced artificial intelligence tools."
Cameron Stanley, Director of Digital Technology and Artificial Intelligence at the US Department of Defense, who spoke at the Palantir conference, announced that the Maven system is "revolutionary." According to him, human involvement boils down to "a left click, a right click, and a left click again."
According to Emelia Probasco, a senior researcher at the Center for Security and Emerging Technologies at Georgetown University, claims about Maven's capabilities may have been exaggerated, and most of the advantage of the United States lay in the volume of incoming data and the skills of the people who used them: "This is not rocket science. I suspect China already has something similar."
In a recent report that analyzed thousands of People's Liberation Army procurement documents, Probasco found that China is building systems that replicate the American ones. For example, China tried to copy the Joint Fires Network, an American program designed to globally combine sensors and weapons so that a drone located on one side of the world could give the command to strike from another drone.
China is clearly leading in some areas. Its dominant position in manufacturing means that it is capable of producing autonomous weapons on a scale unattainable by the Pentagon.
Within the Trump administration, the pursuit of artificial intelligence-based weapons has become almost sacred. Last month, the Pentagon recognized the company as a security threat, partly because it intended to limit the use of its technology in the field of autonomous weapons.
"We will win the artificial intelligence race," Jacob Helberg, Undersecretary of State for Economic Affairs, said last month at the annual Hill & Valley Forum in Washington, which he co—founded with the aim of building links between Silicon Valley and the government.
At the conference, technology company executives, investors, and government representatives were delighted to hear speakers calling on technology companies to provide the military with unlimited access to artificial intelligence.
Palmer Luckey of Anduril argued that building AI capabilities would help prevent major wars. This logic was reminiscent of the Cold War situation: if both sides know what the machines are capable of, neither side will take risks to find out who is stronger.
"Conflicts between superpowers will be mitigated if means are created that are effective enough to deter military action," Lucky said.
However, deterrence presupposes a rational approach, whereas artificial intelligence-based weapons are designed to act faster than the human mind. In the course of research conducted in 2020, experts studied how autonomous systems can accelerate the escalation of conflict and undermine human control, and received rather alarming results.
In one scenario, a system controlled by the United States and Japan responded to a missile launch from North Korea by launching an unexpected retaliatory strike on its own.
"The speed of development of autonomous systems has led to an unintended escalation," according to a report by analysts at the RAND Corporation, a non—profit research organization that cooperates with the military.
General Shanahan, who retired in 2020 and is currently a researcher at the Center for a New American Security think tank, said the race he helped start was haunting him. According to him, governments should set clear boundaries before new technologies get out of their control.
"There is a risk of escalation of the conflict, as a result of which we will have to implement untested, unsafe and unsettled systems if we are not careful, because each side believes that the other is hiding something from it," Shanahan said.
Shira Frenkel is a journalist from San Francisco. Highlights the impact of technology on everyday life and the activities of companies working in the field of social networks.
Paul Mozur is an international technology correspondent for The New York Times. He works in Taipei. Previously, he worked in Hong Kong, Shanghai and Seoul and wrote about technology and politics in Asia.
Adam Satariano is a technology correspondent for The New York Times. He works in London.
