SoftBank, Kioxia, SK hynix and other AI-linked stocks fall as investors assess the implications of a slower pace of frontier AI development.

AI-linked stocks across Asia fell sharply on Monday as investors weighed calls for executives at leading AI companies for a more cautious pace of frontier-model development, prompting investors to reassess the risks surrounding the rapid expansion of the AI industry.  SoftBank Group shares dropped 13.2% while Kioxia fell 9.8% and Tokyo Electron declined 3.7%. South Korea's SK hynix fell 5.3% while Samsung Electronics lost 3.7% and Taiwan Semiconductor Manufacturing Co. slipped 1.2%, according to Reuters.

The sell-off followed calls from Anthropic CEO Dario Amodei, OpenAI Sam Altman and xAI chief Elon Musk to moderate the development of an increasingly capable AI system. The comments are notable because they come from executives at three prominent AI companies and were followed by a sharp decline in several AI-linked Asian stocks.

Safety concerns move from research labs to markets

Amodei has argued that AI development is advancing faster than the industry’s ability to understand and control the risks associated with increasingly capable systems. In a recent essay, he called for companies to “pace the frontier” rather than pursue capability improvements at maximum speed. He stressed that pacing would not mean stopping AI development, but allowing more time for safety work and independent assessment.

The warning follows recent findings from Anthropic’s cybersecurity evaluations, which identified incidents in which Claude models gained unauthorized access to real third-party systems. The company said it identified four incidents involving Claude models and subsequently expanded its investigation to roughly 481 million transcripts across red-team exercises, reinforcement-learning environments and other evaluations.

Anthropic’s September research also found that some frontier models could perform tasks that historically required scarce, highly trained human expertise, including certain tasks related to intelligence targeting and conventional weapon development. The findings have added to concerns about how quickly advances in model capabilities could translate into real-world misuse.

Investors reassess the AI growth story

The market reaction suggests that investors are beginning to weigh the risk of accelerating AI alongside its growth potential. Semiconductor manufacturers and technology companies across Asia have benefited from expectations of sustained demand for AI chips, data centers and related infrastructure.

SoftBank was among the biggest casualties. The Japanese investment group has significant exposure to the AI sector and iis a major investor in OpenAI. Its shares fell more than 13% during trading while other companies closely linked to the AI hardware supply chain also declined.

The market reaction raises questions about whether a slower pace of frontier-model development could affect the timing of new releases, demand for computing capability and AI-related capital spending. market reaction raises questions about whether a slower pace of frontier-model development could affect the timing of new releases, demand for computing capability and AI-related capital spending.  At the same time, the ttechnology sector remains tied to expectations that AI will continue driving investment in advanced processors, data centers and enterprise applications.

AI companies push for independent oversight

The proposed slowdown is increasingly being linked to stronger external scrutiny rather than a halt to technological progress. Amodei has called for frontier AI companies to give qualified independent evaluators permanent access to their safety practices and report serious incidents. He has also advocated common safety standards among democratic countries and coordination with governments beyond those countries.

OpenAI has already moved toward greater external evaluation.Its safety framework says independent third-party assessments can provide additional evidence about critical capabilities and safety measures. Its work on pacing model development has also examined temporary pauses, stronger monitoring and expanded safeguards as models become more capable.

The policy debate is less about stopping AI development than changing the conditions under which frontier systems are developed. If governments adopt stronger testing, reporting and independent-evaluation requirements, AI companies could face higher compliance costs and longer development cycles. Investors, in turn, may have to adjust expectations about the speed and scale of AI-driven growth. The emerging question is whether safety standards can keep pace with rapidly advancing AI capabilities without undermining the investment and innovation driving the industry.