A remarkable debate has erupted in the United States over how fast artificial intelligence should advance. Dario Amodei, CEO of Anthropic, warned recently that AI capabilities could soon advance faster than humans can understand or control them. He called on leading AI companies to “pace the frontier,” slowing capability development enough for safety measures to catch up. His proposal drew support from several prominent figures in the AI industry, including OpenAI CEO Sam Altman, xAI founder Elon Musk and Google DeepMind CEO Demis Hassabis.
US President Donald Trump, however, strongly rejected the idea. His reasoning was straightforward: The US is leading China in AI and must maintain that advantage. “Whoever wins AI wins,” he said. David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, offered another objection. If AI companies believe development is moving dangerously fast, he argued, they can slow themselves down. He also questioned whether their demand for government action was entirely altruistic, suggesting that regulation could help dominant companies protect their market positions.
The debate itself is welcome. AI is developing so rapidly and could have such profound consequences that vigorous discussion is essential. Ethical, scientific, legal, economic and military questions all deserve attention. But the current American debate is somewhat narrowly framed around whether AI should move faster or slower. At least three more fundamental questions need attention.
The first is who should have a voice in governing AI. The current controversy is taking place largely as an American domestic debate. Amodei does address global coordination, but the perspectives of global users and those affected by AI have remained largely missing from the broader debate. This is understandable because American companies lead much of frontier AI development. But America may lead in AI development, while its consequences belong to the world. People around the globe use AI, so their perspectives also deserve a place in the debate. The benefits of AI cross national borders, but so do the dangers. This creates a fundamental imbalance: The risks are distributed globally while much of the decision-making power is concentrated in the United States.
The US government has legitimate concerns about competition with China, but geopolitical rivalry should not overwhelm global safety. Indeed, China exposes the central weakness of any purely American solution. Washington fears that slowing down will allow Beijing to catch up. Beijing has equally strong incentives not to restrain itself if Washington continues accelerating. Each side can therefore use the other’s behavior to justify moving faster. This resembles a classic security dilemma. Even governments that recognize a common danger may intensify it because they fear falling behind their rivals.
The second missing question concerns direction. Speed matters, as Amodei correctly points out. If AI capabilities advance much faster than our ability to understand, regulate and control them, the gap itself becomes dangerous. But speed is not the only issue. Direction matters just as much.
There are at least two kinds of danger. One is human misuse of AI. Criminals could use increasingly powerful AI to conduct cyberattacks, steal money, spread sophisticated disinformation or assist other forms of crime. Military forces could deploy AI-enabled autonomous weapons that identify and attack targets with inadequate human supervision. The other danger is loss of control. An advanced AI system could behave in ways its designers did not anticipate or pursue objectives that conflict with human intentions. In the first case, humans misuse AI. In the second, humans lose control of AI. Both require safeguards.
That is why the real issue is the relationship among speed, direction and human capacity. Slowing development is useful only if safety and control are improved. Moving quickly is acceptable only if safeguards can keep pace. AI needs not only a brake but also a steering wheel.
This leads to the third missing question: What decisions should never be delegated to machines? Developing powerful AI and giving it authority are two different matters. Even if humanity develops extraordinarily capable AI, it does not follow that we should allow it to make every decision. The more powerful AI becomes, the more important it will be to define the boundaries of its authority.
Consider autonomous weapons. Should an AI system be allowed to identify a human target and decide independently to kill? Similar questions arise in nuclear command systems, critical infrastructure, financial markets and medical decisions involving human life. Here, the crucial issue is whether meaningful human control remains. Humans must be able to understand important decisions, intervene when necessary, stop an AI system and ultimately bear responsibility for its actions. Humanity should benefit from AI without surrendering the driver's seat.
The answer is not for countries to race separately but to move forward together. When it comes to AI, we must go together. The US should deepen cooperation with major allies, including South Korea, not only to compete technologically but also to establish common principles for AI safety and human control. Scientists, legal experts, ethicists and security specialists should participate alongside companies and governments.
Amodei is right to call for global coordination, including with China. But international cooperation should go beyond coordination among AI powers and developers. Such broader cooperation need not come at the expense of strategic competition; the two can coexist. Nuclear arms control during the Cold War demonstrated that rivals can establish safeguards even while competing intensely.
The debate started by Amodei is valuable, but “faster or slower” is not the proper choice. AI may be developed first by particular companies and countries, but its future belongs to humanity. The real questions are where we are going, how safely we can get there and which decisions must remain human. The future of AI should not be a race to see who gets there first. We should go together. It should be a collective effort to make sure humanity knows where it is going and remains in the driver's seat.
- - -
Wang Son-taek
Wang Son-taek is an adjunct professor at Sogang University. He is a former diplomatic correspondent at YTN and a former research associate at Yeosijae. The views expressed here are the writer’s own. — Ed.
khnews@heraldcorp.com
