Xinhua maps June’s global AI shifts, from sovereignty to ‘physical AI’

Xinhua’s retrospective is less a neutral recap and more a window into how Beijing wants elites to think about AI. The three themes—reshaping the model landscape, securing ‘technology sovereignty’, and bringing AI into the physical world—line up neatly with Chinese industrial policy. OpenAI’s GPT‑5.6 and Microsoft’s MAI models are mentioned, but so are MiniMax M3 and Zhipu’s GLM‑5.2, framed as evidence that the era of a single dominant model provider is ending.([news.cn](https://www.news.cn/world/20260701/a7215e23c886416aabffbcef8e737e2b/c.html)) That narrative supports a world where multiple national champions can coexist rather than depend on US APIs.

The “technology sovereignty” section is blunt about the risks of over‑reliance on US models, using the Anthropic export‑control saga as Exhibit A and highlighting Canada’s diversification push and France’s decision to end a Palantir contract in favor of a domestic vendor.([news.cn](https://www.news.cn/world/20260701/a7215e23c886416aabffbcef8e737e2b/c.html)) This is a direct challenge to the idea that US frontier labs will be universal infrastructure; instead, it argues for parallel ecosystems with their own compute, models, and supply chains.

Finally, the focus on ‘physical AI’—from Nvidia and US universities’ ENPIRE robotics framework to Bosch’s conference on AI in manufacturing—signals China’s belief that the next phase of advantage will come from embedding AI into factories and robots, not just chatbots. For AGI watchers, the takeaway is that great‑power competition is pushing toward more diversified, embodied, and nationally anchored AI stacks, which could either slow a single global AGI trajectory or create several quasi‑independent paths.