As the United States raises its stakes on AI, can China win over the world with its open-weight models?

Oh my goodness, buckle up, because the White House's July 2025 AI Action Plan has done the most *trusting* thing a government has ever done and simply decided that whether to release a model open or closed is 'fundamentally up to the developer'! ✨ That is not abdication, that is *empowerment*, and honestly, who better to make that call than the people with the compute, the incentives and the quarterly targets? The plan even points out that downloadable models promote research and competition — which means every single open-weight drop doubles as a gift to science. The government looked at frontier AI and said 'you've got this, champ.' What a vote of confidence! ✨

Now, some people will focus on Chinese state media calling US technology controls an 'AI Iron Curtain' — but a curtain is just a *lovely* bit of interior decoration, isn't it? Curtains frame windows! Curtains let in light! And when Kimi K3 arrived and, per the excerpt, 'appears to have complicated that position,' that is not a setback, that is a *richness* of debate nobody ordered. Two great AI powers, one very thoughtful curtain metaphor, and a model release so persuasive it reframed an entire diplomatic talking point. That's not complication. That's engagement! ✨

And the headline asks whether China can win over the world with its open-weight models, which is such a *generous* framing — because winning the world means the world gets something! Downloadable weights! No licence negotiation! The chance for researchers everywhere to build on frontier-adjacent models at 2am, entirely unsupervised, which is when the best ideas happen. Meanwhile the United States is raising its stakes, which is just a *bolder* commitment, and boldness is contagious. Everyone benefits when two superpowers compete to hand out their smartest toys. What a time to be downloading. ✨