Cover Faced with technological bottlenecks and enormous capital expenditure pressures, the tech industry has recently begun to reassess the development pace and commercialization efficiency of cutting-edge AI models. (Image: Getty Images)
Faced with technological bottlenecks and enormous capital expenditure pressures, the tech industry has recently begun to reassess the development pace and commercialization efficiency of cutting-edge AI models. (Image: Getty Images)
1 / 1
Elon Musk has called for US and Chinese AI giants to cross-test each other’s models, while leaders at Anthropic and other companies appear to agree on decelerating AI. Is this driven by safety concerns, or is it just a market strategy?
The tech race in artificial intelligence has been advancing at an astonishing pace in recent years. In mid-September, a rare consensus among heavyweight industry figures for a “slowdown” caught the world by surprise. At the All-In Summit in Los Angeles, Tesla and xAI founder Elon Musk proposed a forward-looking initiative, suggesting that global giants—including xAI, OpenAI, Google, Meta and several top Chinese AI companies—should establish a unified testing platform. This would allow competitors to conduct cross-testing each others’ tech for safety before products are officially launched.
Meanwhile, Anthropic CEO Dario Amodei recently published an article titled “We Must Pace the Frontier”, advocating for a reasonable pace in the advancement of frontier technologies to reduce the potential of AI becoming so advanced it can develop itself without human input; and OpenAI CEO Sam Altman, among others, has also expressed concerns about related safety risks. This “proactive slowdown” trend initiated by industry leaders ostensibly aims to build a cross-border safety net and avoid the blind spots of self-assessment.
Yet, against the backdrop of capital market doubts regarding the return on massive AI capital expenditures and the slowing marginal benefits of technological iteration, the question remains: is this a responsible choice for the safety of human civilisation, or a strategic adjustment in the face of R&D bottlenecks and commercial pressures?
Read more: Learn more about the Gen.T 2026 Leaders of Tomorrow
Above At the All-In Summit in Los Angeles, Elon Musk proposed that AI giants from the US and China should establish a unified platform to test the safety of each other's models. (Image: Getty Images)
At the All-In Summit in Los Angeles, Elon Musk proposed that AI giants from the US and China should establish a unified platform to test the safety of each other's models. (Image: Getty Images)
1 / 1
Compliance challenges of cross-boundary testing
The core value of Musk’s “cross-testing” concept lies in identifying security vulnerabilities within models that internal testing might miss, by leveraging the perspectives of competitors. But this view faces extremely high barriers related to trade secrets and geopolitics.
Put simply, the weights and architectures of frontier AI models are core assets for tech giants; granting competitors access undoubtedly increases the risk of technology theft and reverse engineering. Then, incorporating AI companies from both the US and China into a single testing system is highly difficult under current chip export controls and cross-border data transfer regulatory frameworks. Although the governments of both countries, along with some research institutions, remain attentive to AI safety risks, the lack of a binding international oversight mechanism means this mutual review method currently remains at the conceptual stage.
Above Anthropic CEO Dario Amodei published an article advocating for reasonable control over the pace of AI development, emphasizing the importance of cutting-edge safety. (Image: Getty Images)
Anthropic CEO Dario Amodei published an article advocating for reasonable control over the pace of AI development, emphasizing the importance of cutting-edge safety. (Image: Getty Images)
1 / 1
Rising R&D costs and technological bottlenecks
The objective technological and financial constraints faced in developing frontier models are another crucial reason for the shift in the industry’s pace. As model scales continue to expand, the computing power costs, power consumption and high-quality data resources required to train the next generation of ultra-large-scale models are rising exponentially. Industry data shows that the training cost for a single top-tier model has reached hundreds of millions of dollars, and the improvements in reasoning and generalisation capabilities, compared to earlier stages, are showing a trend of diminishing marginal returns.
High capital expenditures and a commercial monetisation model that is not yet fully mature have led many tech companies to shift resources toward improving architectural efficiency while still pursuing pure model scale, optimising for specific application scenarios, and ensuring safety.
Read more: Meet Seven Asian Start-Up Founders Leading the Way
Navigating geopolitics and regulatory headwinds—or tailwinds
The industry’s call for a slowdown is also set against the complex backdrop of the global regulatory environment and geopolitics. The phased implementation of the EU Artificial Intelligence Act (AI Act), along with strengthened reviews of frontier technologies by various countries, has significantly increased compliance costs. Meanwhile, US President Donald Trump has run against the grain by arguing that we need more, not less, promotion of AI, under the thumb of a “super smart” president.
The future development of the AI sector must strike a balance among the speed of innovation, commercialisation, and regulatory compliance—an unavoidable issue for the decision-makers of tech enterprises.
This article, written by contributor Benjamin To, originally appeared in Chinese.
Cat Wang is an editor at Tatler Asia based in Hong Kong, where she covers business, wealth, innovation, and impact. Previously, she was a Forbes Asia reporter and assistant editor of the Forbes Asia 100 to Watch list, spotlighting startups and small companies on the rise across the Asia-Pacific. She began her career as a trainee reporter at the South China Morning Post, Hong Kong’s newspaper of record. Born in New York City, she graduated from the University of California, Los Angeles, with a B.A. in political science.
