AI Library
Books for Reading AI
Choose a book, then read it in order from the table of contents.
[AI Library] Chapter 35. Arms Competition and International Norms: The Dilemma of Export Controls and Technical Standards
Chapter 35. Arms Competition and International Norms: The Dilemma of Export Controls and Technical Standards
Arms Race and International Norms: Dilemma of Export Controls and Technology Standards In the early morning of July 16, 1945, the first nuclear explosion in history occurred in the desert of New Mexico, USA. Robert Oppenheimer, head of the Manhattan Project, looked at the flash and recalled a passage from a Hindu scripture: “I am now death, the destroyer of worlds.” Humanity opened Pandora's box that day. And now, we are opening the second box. It's an AI arms race. Russian President Vladimir Putin said in 2017: “Whoever controls artificial intelligence will rule the world.” These are not empty words.
China is making all-out efforts to become the world's No. 1 in AI by 2030. The United States has pledged to field thousands of autonomous weapons systems on the battlefield within 18 to 24 months through its 'Replicator' initiative. This is not just a technology race. It is a war where the survival of the nation depends. The problem is that there are no breaks in this competition. If the United States develops, China must also develop. If China develops, Russia must also develop. “The enemy has an army of robots, so will we be the only humans fighting?” This fear is sweeping the world.
International norms and controls are falling behind, unable to keep up with the pace of technology. Semiconductor War, Oil Embargo of the 21st Century During World War II, the United States banned oil exports to Japan. Without oil, tanks, fighter jets, and warships cannot move. Japan advanced south to secure oil and eventually attacked Pearl Harbor. Oil changed the course of the war. Today's oil is a semiconductor. To train and operate AI, a high-performance GPU, or graphics processing unit, is essential. Chips like NVIDIA's A100 and H100 are no longer just computer parts. It is a strategic material.
Without these chips, large-scale AI models cannot be trained. You can't create two brains for an autonomous drone. The United States is enforcing strong export controls to curb China's military AI capabilities. The semiconductor export regulations to China, which began in 2022, are being strengthened every year. In October 2023, the U.S. Department of Commerce's Bureau of Industry and Security significantly strengthened control of high-performance AI chips. The chip's total computational power, performance density, and data
Control targets are identified based on three criteria: whether they are designed for a specific purpose; The strategy is to physically prevent China from developing advanced military AI. But this is like building a sand castle. Water always finds a way. Several semiconductor companies circumvented regulations by selling low-performance AI chips in the Chinese market. When the United States lowers its acceptance standards, companies design chips with performance just below the standards. The cat-and-mouse battle of regulations and detours continues. There is a bigger problem. Hardware may be held at customs. Because it's a physical object.
But what about software? The algorithm crosses borders on a USB stick or in an encrypted email. It is impossible to completely control open source AI models shared over the Internet. There are reports that Chinese research labs are already using the U.S. open source model to analyze military intelligence. There is an even bigger nightmare: the proliferation of low-cost AI weapons, a poor man's air force. It is not a competition between great powers. Proliferation. In the past, precision-guided weapons and stealth aircraft were the preserve of a few powerful countries.
It required trillions of won in development costs and decades of technology accumulation. However, AI technology has dramatically lowered the barrier to entry. In the war in Ukraine, we saw $500 commercial drones equipped with simple AI software to destroy multi-million dollar tanks. This is the maximization of asymmetric power. As of 2025, the Russian military is launching large-scale drone attacks on Ukraine. Hundreds of drones fly in one airstrike. Between September 9 and 10, 800 attack drones were deployed. About 20 of them entered Polish airspace, and NATO defenses only shot down four of them.
In the face of large-scale drone attacks, the limitations of traditional air defense systems have been revealed. Now terrorists and criminal organizations have also become dangerous. With just graduate student-level coding skills and a commercially available drone, you can create an autonomous lethal weapon that uses facial recognition technology to assassinate specific people or attack crowds of people. This is the era of ‘DIY killer robots’. It is difficult to identify the perpetrator of the attack, and it can cause uncontrollable damage to civilians. As technology becomes more democratized, so too does fear.
Dilemma of international norms: ban or manage?
The international community recognizes this risk. The UN Convention on Certain Conventional Weapons (CCW) has been discussing regulations on autonomous lethal weapons (LAWS) since 2014. However, even after 11 years, an agreement is still far away. In December 2024, the United Nations General Assembly adopted a resolution on autonomous weapons of destruction. 166 countries were in favor, 3 countries (Russia, Belarus, and North Korea) were against it, and 15 countries abstained. The resolution mentioned the possibility of a “dual approach” that would ban some autonomous weapons and regulate others.
UN Secretary-General António Guterres has called for a binding agreement to ban autonomous weapons operating without human control by 2026. But the reality is different. Military powers such as the United States, Russia and Israel oppose a blanket ban. Their logic is this: “AI can strike more accurately than humans and reduce civilian casualties.” In fact, if you look closely, it is because they do not want to give up the advanced weapons that are already under development. Human rights groups and some countries claim otherwise. “Machines should not be allowed to decide human life or death.” This is ethically sound.
Small island states like Kiribati advocate for a complete ban. The country underwent 33 nuclear tests between 1957 and 1962. I know better than anyone the horrors of new weapons. South Korea is taking a pragmatic stance. We are opposed to a total ban on LAWS, and we must ensure 'Meaningful Human Control'. In the face of the existential threat of North Korea and the reality of a population cliff, AI weapons are a means of survival that we cannot give up. CCW meetings are run by consensus. Opposition from just one country could derail the proposal. That is why there has been no real progress for over 10 years.
This is why creating norms is so difficult. The war of technical standards, who sets the rules. Along with norms, what is important are technical standards. The future battlefield is a place where all weapons are connected through a network. For the US's JADC2 (Joint Domain Command and Control) or Korea's manned and unmanned complex systems to succeed, a common language and rules are needed for different AI systems to communicate. The question is who sets these standards. America wants the world to follow its standards. NATO members must follow U.S. standards to be compatible with U.S. equipment. NATO in 2025
In announcing the data utilization strategy, interoperability and integration were emphasized. This is not just a technical issue. It's a question of technology dependence. Countries with independent defense production capabilities, such as Europe and Korea, are worried. Following American standards makes joint operations easier. But at the same time, it is dependent on American technology. Israel is building its own AI ecosystem and calling for ‘sovereign AI’. The standard dilemma is this. Stronger standards lead to better safety and interoperability.
Instead, the pace of innovation slows, and the country that leads the standards sets the rules. When standards are weak, individual development becomes faster. Instead, communication between allies on the battlefield is not possible. The moment there is no connection, AI becomes an isolated toy rather than a power multiplier. The world is being divided into two blocks by the AI technology barrier. The standards of the United States and its allies, and the standards of China and its allies. Like the Iron Curtain during the Cold War, a digital curtain divides the world.
In front of Pandora's box, we are now in a similar position to the nuclear weapons developers in 1945. The technologies we create can either make the world safer or lead to destruction. The AI arms race is an unavoidable reality. Because our enemies will not stop just because we stop. However, indiscriminate competition leads to mutual destruction. We must compete smartly. First, reliability must be secured. We must be able to control our own AI. Investing in explainable AI (XAI) and security technologies is a necessity, not an option. Like DARPA's SABER program, we need to create a robust AI that is not shaken by enemy attacks.
Second, international cooperation is needed. At the very least, export controls and proliferation prevention systems must be strengthened to prevent lethal AI technology from being passed on to terrorists or rogue states. Although complete control is impossible, you can slow it down. Third, human control must be maintained. We must adhere to the 'Human-in-the-loop' principle, which leaves the authority to pull the trigger to humans, not machines. This is a matter of ethics and at the same time a matter of survival. Weapons that we cannot control will ultimately harm us.
As a fighter pilot, you want better weapons. But you don't want a weapon you can't control. Even in the era of algorithmic warfare, the responsibility and ethics of war must ultimately remain the responsibility of humans. That's what makes us different from machines. That is the final line we must keep. From the cockpit, international norms feel as light as paper. However, as the paper accumulates, the direction of weapons development and the market changes. In the era of algorithmic warfare, not only missiles but also standards, licenses, and regulations fly in formation. If you don't know that, you'll be late on the battlefield.
If you're late, it's over.
Part 8. Future Outlook and South Korea’s Choice
Kim Kyung-jin
Attorney · Former Member of the National Assembly · AI Policy Researcher
© 2026 Kim Kyung-jin. All rights reserved.