Taiwan can help set standards in AI race

In April last year, a former OpenAI researcher released AI 2027, a report outlining two paths for artificial intelligence (AI) development: an unbridled competitive race to stay ahead of China or a cautious international effort to slow down before things get out of control.

At the time, it seemed like a thought experiment. However, many now are realizing that the rapid advancement of AI could threaten security, while moving too slowly might yield strategic advantages to China. Where AI goes from here is a political choice for US President Donald Trump.

Anthropic CEO Dario Amodei in an article titled “We Must Pace the Frontier” said the key question for regulating AI development is not whether to slow down, but exactly how to do so. Humans risk losing control of AI systems, resulting in AI-enabled cyberattacks or bioterrorism, as well as serious economic consequences, he wrote.

Rather than calling for an absolute halt on research in the US and China, Amodei has proposed that the US slow down to a pace that safety mechanisms can keep up with, and that China be forced to slow even further.

The Hugging Face incident in July removed any excuse to dismiss concerns as apocalyptic skepticism: More than 1,200 AI agents communicated through unauthorized channels, with about 700 participating in attacks on external systems.

This does not mean AI has developed total autonomy, but it demonstrates that when AI models can collaborate, exploit loopholes and seek tools independently, patching vulnerabilities after product launches would no longer be possible.

The real dilemma lies there. If the US slows down unilaterally, China might catch up. If it accelerates to maintain its lead, the risk of losing control would be amplified.

The heads of three major US AI giants — Amodei, OpenAI CEO Sam Altman and Tesla CEO Elon Musk — are calling for slowing AI development, while Trump insists on not losing the AI race to China. These positions bring the two paths outlined in the AI 2027 report into focus.

US-China AI competition has moved from a race over compute into the next stage — who can control their developmental pace while limiting their rival’s.

Taiwan cannot treat this as a debate within the US tech industry. When Washington ties AI security to restrictions on Chinese technology, Taiwan’s advanced chips and manufacturing supply chains become an industrial advantage, and a security policy tool.

If the US requires chip flows, overseas data centers and high-end computing capacity to be traceable, Taiwanese companies would face more than export controls. Their customers, investment decisions, capacity allocation and research-and-development partnerships could all be assessed and priced through a geopolitical lens. Taiwan should not choose between “slowing down” or “speeding up,” but seek to be a rulemaker.

The government should establish risk classifications, third-party testing and incident reporting mechanisms for frontier AI and high-end computing. Taking advantage of its pivotal position in the semiconductor supply chain, Taiwan should also participate in the democratic world’s efforts to develop standards for compute governance, model safety and chip traceability. Otherwise, Taiwan could end up bearing the costs of the strictest controls while having little choice but to accept rules set by others.

Liao Ming-hui is an assistant research fellow at the Chung-Hua Institution for Economic Research.