중국의 오픈웨이트 AI 모델은 BRICS 국가들(Brazil, Russia, India, China, South Africa)에 서방 플랫폼의 대안으로 공급하도록 준비되어 있다.
At the BRICS summit in New Delhi, Chinese President Xi Jinping announced an initiative to help the group's eleven member states build a shared open-source AI base—precisely as the United States grapples with whether to slow development of its most capable models. Anthropic CEO Dario Amodei, who proposed such a slowdown, identifies China as the "biggest open question" in his framework.
China is positioned to make this offer credible. It currently leads in open-weight models, a category spanning Moonshot AI's Kimi K3 (the largest open-weight model released to date), DeepSeek, Zhipu's GLM series, MiniMax, and corporate players including Xiaomi and Alibaba. Alibaba's Qwen family has become the default base for countless fine-tunes worldwide, and some recent models train on Huawei chips rather than Nvidia hardware. For countries lacking gigawatt-scale data centers, freely downloadable weights represent the fastest path to indigenous AI infrastructure—the precise gap Beijing targets with its BRICS initiative.
Xi presented five initiatives, the first an "open source and inclusive AI initiative." China proposes to pioneer a BRICS AI open-source community, support large language model development and deployment, conduct specialized AI seminars and training, and build an open AI ecosystem. The others address trade facilitation, a cloud platform for digital infrastructure, smart manufacturing standards, and an engineer training alliance with mutual competency recognition.
Xi framed this within a broader speech on rule-setting, arguing that international rules must be written by all countries and that "might makes right" has no place in modern governance. He called for WTO reform, greater developing-country weight in international financial architecture, and a consensus-based global AI governance framework enabling new technologies to "light the way toward shared prosperity." BRICS now comprises Brazil, China, Egypt, Ethiopia, India, Indonesia, Iran, Russia, Saudi Arabia, South Africa, and the United Arab Emirates.
Yet China's commitment to true openness remains unclear. The Ministry of Commerce is reportedly weighing export controls on AI models with tiered requirements: filing for less capable systems, security reviews for stronger ones, and potential bans on public release for the most capable. Consultations with Alibaba, ByteDance, and Zhipu are ongoing. Xi's speech left unresolved how a BRICS open-source community would operate under such restrictions, and what funding, timeline, and specific models would be involved.
The United States pursues an opposite argument—about slowing rather than opening. Amodei recently published "We Must Pace the Frontier," proposing independent auditors with deep lab access, shared safety standards among leading developers in democratic countries, and eventually international agreements covering dangerous applications. OpenAI CEO Sam Altman pledged external auditor access "comparable to that of employees," and Elon Musk briefly endorsed the plan. For Amodei, geopolitics is the hardest piece. China represents the "toughest dilemma," he said on CBS "Sunday Morning," because a shared speed limit faces enormous incentives to defect—"the military advantage that you get from that are so large." He stressed: "I don't know if it's possible, but we should try." His framework pairs slower development with tighter chip export controls and better theft prevention.
The White House disagreed. During a stop in Ireland, U.S. President Donald Trump dismissed the appeal: "We're leading China in AI, and frankly, I want to keep it that way, because whoever wins AI, wins." He considers basic safeguards reasonable but dismisses warnings of catastrophic risk as overblown.
The two tracks converge soon: Xi is expected in the United States later this month, with AI governance likely a central topic in his talks with Trump.