Right after Moonshot AI stunned the tech community by releasing an open-source model only marginally behind the American frontier, Chinese Premier Xi Jinping has indicated that China will continue to promote open-source AI.
Speaking at the opening ceremony of the World Artificial Intelligence Conference (WAIC) in Shanghai, Xi called for adhering to the principles of “openness” and “win-win,” underscoring the need to encourage open sourcing and collaboration in AI development. The timing is hard to ignore. Just a day earlier, Beijing-based Moonshot AI had released Kimi K3, a 2.8-trillion-parameter model that the company says is the largest open-source model released to date, and one that placed third on the Artificial Analysis Intelligence Index, trailing only Claude Fable 5 and GPT-5.6 Sol. Moonshot has said it will open-source the model’s weights within days, making a system that competes with the best proprietary labs in the world available to anyone with the hardware to run it.
Xi’s address leaned heavily on his familiar framing that the world is moving through “great changes unseen in a century.” He told the gathered delegates that AI development has entered “an unprecedented period of active innovation,” and that this carries “great opportunities as well as challenges to governance.” It’s the kind of line Xi has used before to describe shifts in global power, and its presence at the top of an AI conference speech signals how central Beijing now considers the technology to be in that broader realignment.
He also used the platform to take a swipe at how AI and national security are increasingly being talked about in the same breath, particularly in Washington. Xi warned against “generalising” the concept of national security, and against one country prioritising its own security interests over those of others in the AI field. China opposes these practices, he said, adding that “new historical injustices” should not be created around AI — among the more pointed lines of the speech, and a fairly clear, if indirect, reference to the export controls that have shaped the US-China AI rivalry for years.
The speech came just hours after Donald Trump delivered a prime-time address in which he accused China of interfering in the 2020 presidential election, and on a day when US officials confirmed that Nvidia had shipped only a handful of H200 chips to China despite the White House easing restrictions on a case-by-case basis back in December. Against that backdrop, Xi’s emphasis on open-source AI reads as much like a geopolitical position as a technological one. Where the US has largely kept its most capable models closed and behind APIs, China’s leading labs — Moonshot, DeepSeek, Alibaba, Z.AI — have made open weights their calling card, and Kimi K3’s benchmark performance gives Beijing a genuine case to point to when it argues that openness doesn’t have to come at the cost of capability.
Xi also announced that China will provide 5,000 AI research projects to developing countries over the next five years, alongside training programmes, seminars and cooperation centres, extended to members of ASEAN, the Arab League, the Shanghai Cooperation Organisation, the African Union, Latin American and Caribbean states, and the BRICS bloc. He framed this as part of China’s responsibility as a “major country” to provide what he called international public goods in AI, noting that the country’s core AI industries have already crossed 1 trillion yuan, or roughly $148 billion.
The speech also referenced the World AI Cooperation Organisation, which was formally established in Shanghai a day earlier and has so far drawn 29 signatory countries, including Russia and Indonesia. Xi described AI’s development as something that should not be a “solo performance” by any single nation, but rather a “symphony of global collaboration” — language that positions the new body as a counterweight to US-led AI governance efforts, most of which China has stayed outside of.
Whether the open-source strategy actually widens China’s influence over how AI gets built and deployed globally will depend on developers outside China choosing to build on models like Kimi K3 over closed alternatives from OpenAI, Anthropic and Google. But with a model that trades blows with the best proprietary systems now sitting in the open, and a head of state making the case for openness on stage right after, Beijing’s pitch is at least a coherent one this time around.