Asia is reaching for the open road that Chinese AI offers
At a recent Apec digital technology and AI forum in Chengdu, ministers across the Asia-Pacific Economic Cooperation forum issued a statement in support of open-source artificial intelligence models – a rare point of agreement in a debate often framed as a zero-sum US-China contest. It pointed towards an AI order that cuts against the assumption of a contest with a single winner, an assumption…
At a recent APEC digital technology and AI forum in Chengdu, ministers from the Asia-Pacific Economic Cooperation forum endorsed open-source artificial intelligence models. This agreement stands out amid debates often framed as a zero-sum US-China competition. The statement suggests an AI order that counters the assumption of a single winner, a notion still influencing Washington's concerns about Chinese AI advancements.
Across Asia, governments are constructing their own systems, with most relying on open Chinese models. This openness fosters a multipolar AI landscape, where nations control their capabilities instead of relying on foreign platforms.
Three distinct paths are emerging in this AI order: the flagship national model built on a Chinese base, government pilots integrating Chinese systems into public services, and private startups utilizing the most open models. China's growing influence stems from its alignment with these goals, rather than exceeding Washington's efforts.
The economics support this shift. Creating a frontier model is costly and requires rare talent, leaving only China and the United States at the forefront. For others, an open Chinese model eliminates barriers: the weights are free, run on local servers without data export, and cost 60-90% less to operate than leading American systems, only trailing by a few months.
The flagship national model is the most visible path. Singapore moved to Alibaba's Qwen for its SEA-LION project, while Kazakhstan's AlemLLM was developed by Chinese startup 01.AI using DeepSeek. Despite relying on China's base, these models remain adaptable and controlled by the respective governments.
Government pilots are another route. Malaysia's Prime Minister's Office and communications ministry were early adopters of a domestic platform running DeepSeek. Vietnam recently passed a law asserting sovereignty over its AI infrastructure. These pilots foster domestic expertise and options no single supplier can revoke.
Private startups reach beyond developing nations. In Japan, Abeja and Lightblue build models on Qwen, surpassing American systems in Japanese-language tasks. South Korea, where DeepSeek faces security restrictions, turns to Qwen due to its permissive licensing. In Kenya and Nigeria, startups leverage Chinese models for up to 90% lower inference costs.
In essence, this does not represent Chinese domination. As Asia adopts open models, shared tools and standards spread, granting China more influence by aligning with this vision. The US, however, continues to fuel concerns by treating frontier models as licensed capabilities that can be withdrawn. For many in the region, the sovereign choice has already been made, following an open road that incorporates China.
Written by urgent.news from South China Morning Post's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
