SEOUL, September 14 (AJP) - After years of acceleration and frenzied infatuation with generative artificial intelligence, U.S. creators and users are suddenly shouting time-out, with debate spreading among incumbent and former presidents, Silicon Valley and beyond over warnings of a Frankenstein moment and even doomsday for humanity.
The sober reckoning, raised most dramatically by former OpenAI and Anthropic researchers, has arrived just as the latest generation of models pushes closer to artificial general intelligence, or AGI.
Even the people running the race are beginning to question its speed.
Anthropic Chief Executive Dario Amodei, after years of competing fiercely with OpenAI and other frontier laboratories, called for greater restraint.
"Over the last few months, I have become convinced that fully addressing the risks requires even more prudence—not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up," Amodei wrote Saturday.
OpenAI Chief Executive Sam Altman quickly backed stronger independent evaluation. Elon Musk of xAI responded, "Dario is right," while Google DeepMind Chief Executive Demis Hassabis also lent his support.
Their convergence is striking because all four are competing to build increasingly powerful AI.
U.S. President Donald Trump, whose administration has put beating China at the center of its AI strategy, was clearly unhappy with the interruption.
Trump said the United States was leading China and that "whoever wins AI wins."
"We can put guardrails, we can do this and that, but I think you have a lot of negative forces that are bringing it up that ... shouldn't be bringing it up, and they're bringing up things that won't happen," Trump told reporters Sunday during a weekend trip to Ireland.
Former President Barack Obama took almost the opposite position at a closed-door Democratic fundraiser in New York on Sept. 10.
The technology is "moving very fast in private hands," he said, and could become dangerous if it is not brought under control.
For South Korea, however, the argument is no philosophical exercise.
Seoul is only now preparing its biggest push into AI after starting late against the United States and China. Slowing down could widen that gap. Racing ahead without stronger safeguards could leave Korea repeating mistakes that the companies at the technological frontier are only beginning to confront.
South Korea's 2027 budget bill raises spending on AI and three related flagship industrial projects by 97.2 percent to 21.3 trillion won ($15.8 billion).
The National AI Computing Center planned for Solasido is targeting 15,000 GPUs by 2028, while Nvidia agreed late last year to supply 260,000 GPUs to Korean buyers.
Korea, in other words, is preparing to sprint at precisely the moment some of the world's fastest runners are asking whether the race itself has become dangerous.
Countries already running behind can hardly afford a breather, much less a moral reconsideration, if their competitors keep moving.
China has continued rolling out increasingly capable models, while Washington increasingly treats AI leadership as a matter of economic and national security.
"The government should hold its own center of gravity and not be rattled by a single remark from Trump," said Moon Hyoung-nam, professor of interdisciplinary studies at Sookmyung Women's University.
"The United States has spent 2026 institutionalizing this, down to fast-tracking AI for national security and locking in AI talent," Moon said. "Korea needs to execute infrastructure, talent and industrial application at the same time, rather than issue declarations."
AI debate in Korea has largely centered on catching up — computing capacity, sovereign models, data centers, talent and industrial adoption — rather than on what happens when increasingly autonomous systems become difficult to control.
It has institutionalized risk management now being debated in Silicon Valley.
The AI Basic Act, in force since Jan. 22, requires lifecycle risk management for systems meeting three conditions simultaneously, including cumulative training compute above 10^26 floating-point operations.
No known domestic developer currently reaches that threshold.
Korea has introduced rules for frontier-scale AI before producing much frontier-scale AI of its own, while massive public and private investment is rapidly expanding the computing infrastructure that could eventually bring Korean systems within reach of those rules.
But the casualty has become visible in Korea's youth job market.
Youth employment fell by 285,000 between June 2022 and June 2026, and industries with high exposure to AI accounted for 268,000, or 94 percent, of the decline, according to a Bank of Korea study released this month.
Employment among young people fell 31.4 percent in information services, 27.4 percent in publishing, 16.6 percent in computer programming and 11.6 percent in professional services.
The question is therefore not whether Korea should abandon its AI push.
It is whether safety, cybersecurity and evaluation can be built into the expansion before Korea reaches the scale at which failures become harder and more expensive to contain and whether the country can absorb the labor-market disruption already emerging as AI begins to erode the entry-level jobs.
Some Korean experts acknowledge the trade-off to some extent.
"It is like stretching chewed gum. Holes open up in the middle."
The answer is not to cut investment, Kong said, but to change strategy.
"Securing a general-purpose model is close to intractable, because it has to stay open to everyone," Kong said.
"Korea could build enough AI usage cases through actual security attack and defense training, and we just might be able to stand on equal terms with the United States."
Korea cannot realistically stop running while Washington and Beijing are still way ahead in the game.
It also has little reason to reproduce an American development model whose own architects increasingly say has allowed capabilities to run ahead of safeguards.
For Korea, the immediate danger is not necessarily Skynet.
It is arriving late to the AI race and then running so hard to catch up that it ignores the warning signs being raised by those already ahead.
Trump, asked whether the industry should slow down or face greater regulation, allowed that some rules might eventually be needed but named none.
He answered after watching a golfer tee off.
AJP Takeaways
- South Korea is sharply accelerating its AI investment just as leading U.S. researchers and executives are warning that capability development may be outrunning safety and control.
- The 2027 budget allocates 21.3 trillion won ($15.8 billion) to AI and three related flagship projects, while Korea is rapidly expanding computing capacity through the National AI Computing Center and large-scale GPU purchases.
- Korea already has frontier-AI risk rules, but no known domestic developer currently meets the legal threshold, giving Seoul a window to strengthen security and evaluation before domestic systems reach much greater scale.
- The Korean dilemma is not simply whether to slow down. Falling further behind the United States and China carries strategic costs, but racing ahead without stronger safeguards could import the same problems Silicon Valley is now confronting.
Copyright ⓒ Aju Press All rights reserved.