Tertius News is an AI-native newsroom: an AI model reads the source articles linked below — on this story, all of them from one outlet — and extracts what that outlet reported, claim by claim. How this works →

New information became available — this story may be updated shortly.

China urged to avoid ‘us or them’ split with US over AI governance

At an APEC digital technology and AI forum in Chengdu, ministers issued a statement supporting open-source AI models, a rare consensus in a field often framed as a US-China contest. Across Asia, governments and start-ups are building on open Chinese systems, attracted by lower costs and local data control.

By Tertius News AI Desk 1 outlet · 1 article Version 2 Coverage Aug 23, 2026 Published 2d ago

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, according to the South China Morning Post. The statement was described as a rare point of agreement in a debate often framed as a zero-sum US-China contest, pointing towards an AI order that cuts against the assumption of a contest with a single winner.

Openness driving adoption

The South China Morning Post reports that leading Chinese AI systems are open weight, free to download and reshape, which is producing an AI order in which no single company or country owns the field. This openness matches what many states pursuing sovereign AI want: a multipolar landscape where they control their own capabilities rather than rent them from a foreign platform.

The economics are a significant factor. Building a frontier model costs billions of dollars and demands rare talent, leaving just China and the United States still training at the frontier. For everyone else, an open Chinese model removes the barrier: the weights are free, they run on local servers without sending data abroad, and they cost 60 to 90 percent less to operate than the leading American systems while trailing by only months.

Three paths to adoption

The South China Morning Post outlines three distinct roads countries are taking to enter this order: the flagship national model built directly on a Chinese base, the government pilot that puts Chinese systems into public services, and the private start-up that ships on whatever is most open.

The most visible is the flagship national model. Last year, AI Singapore moved from using Meta's Llama in its Southeast Asian Languages in One Network (SEA-LION) project to running SEA-Lion large-language models on Qwen by Alibaba Group Holding, according to the report.

In another example, Kazakhstan's AlemLLM was developed by Chinese start-up 01.AI on a DeepSeek base. The report says this was the only option for a country with only a few hundred ageing chips.

The report also notes the World AI Cooperation Organisation (Waico) was signed into existence in Shanghai at the World Artificial Intelligence Conference in July, with an initial 29 members, mainly from the Global South.

How this outlet told it

South China Morning Post as of Aug 23, 2026, 8:30 AM UTC

Framing: Emphasizes the positive potential of Chinese AI openness for Asia, framing it as an 'open road'. — Optimistic and analytical; supportive of multipolar AI development.

Facts Included:

At a recent APEC digital and AI forum in Chengdu, ministers across the Asia-Pacific Economic Cooperation forum issued a statement in support of open-source AI models.
The statement was described as a rare point of agreement in a debate often framed as a US-China contest.
AI Singapore moved from using Meta's Llama in SEA-LION to running SEA-Lion large-language models on Qwen by Alibaba Group Holding.
Kazakhstan's AlemLLM was developed by Chinese start-up 01.AI on a DeepSeek base, the only option for a country with only a few hundred ageing chips.
Open-weight Chinese AI systems cost 60 to 90 percent less to operate than leading American systems.
Chinese AI systems allow countries to run on local servers without sending data abroad.

Read the source article →

What was claimed, and who backed it

Each row is one claim, attributed to the outlet whose wording states it most clearly. Confidence rates how directly the source text states the claim — explicit and unhedged rates high; hedged, pieced-together, or internally inconsistent statements rate lower. It does not measure whether the claim is true. Status counts the distinct outlets we found asserting it — so a single-source claim can still show high confidence, and a multi-source claim can show medium. Every one of those outlets is named beside the status, so you can check the count against the list. For claims extracted before we began storing that list, the row says so: it names the outlet the claim is quoted from and states that we have not recorded which outlets backed it. Outlets wrote at different times, so a figure that evolves — a casualty count, for example — can legitimately differ between rows; check the "as of" time next to each claim's source.

Claim	Confidence	Status
ClaimAt a recent APEC digital technology and AI forum in Chengdu, ministers issued a statement in support of open-source artificial intelligence models.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimThe APEC statement in support of open-source AI models was a rare point of agreement in a debate often framed as a zero-sum US-China contest.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimAI Singapore moved from using Meta's Llama to using Qwen-based large-language models for its Southeast Asian Languages in One Network (SEA-LION) initiative.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimKazakhstan's SEA-LION model was developed by Chinese start-up 01.AI on Alibaba's Qwen base, the only option for a country with only a few hundred ageing chips.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimOpen-weight Chinese AI systems cost 60 to 90 percent less to operate than leading American systems.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimOpen Chinese AI models allow countries to run AI on local servers without sending data abroad.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimAI Singapore moved from using Meta's Llama in its SEA-LION project to running SEA-Lion large-language models on Qwen by Alibaba Group Holding.	ConfidenceHigh	StatusSingle-source via South China Morning Post as of Aug 23, 2026, 8:30 AM UTC
ClaimThe World AI Cooperation Organisation (Waico) was signed into existence in Shanghai at the World Artificial Intelligence Conference in July, with an initial 29 members, mainly from the Global South.	ConfidenceHigh	StatusSingle-source via The Next Web as of Aug 19, 2026, 9:30 AM UTC

Transparency & Audit Trail

1 outlet · 1 article consulted: South China Morning Post 8 claims extracted Version 2 Written 2026-08-23