BEIJING, March 23, 2026, 21:39 CST
- Chinese open-source AI keeps gaining ground, carving out what the U.S.-China Economic and Security Review Commission calls a “self-reinforcing competitive advantage” even as Washington tightens chip restrictions. 1
- Tencent hooked OpenClaw up to WeChat on Sunday; Alibaba followed with a fresh agent platform rollout Monday. Chinese tech giants aren’t wasting time plugging AI agents into their flagship apps. 2 3
- Siemens has noted that some of its developers are turning to Chinese AI models for industrial projects, citing lower costs and simpler tuning as key reasons. 4 5
China’s open-source AI drive is taking direct aim at U.S. tech dominance, the U.S.-China Economic and Security Review Commission said Monday, flagging signs that lower-cost Chinese AI models are carving out a “self-reinforcing competitive advantage”—even after years of U.S. restrictions on chip sales. According to the commission, Chinese models now crowd the top spots on developer platforms like HuggingFace and OpenRouter. 6
The warning lands as Chinese tech firms scramble to commercialize surging model buzz. Tencent plugged OpenClaw into WeChat on Sunday. Alibaba followed with Accio Work on Monday, hot on the heels of last week’s Wukong debut—another step as agentic AI, which handles multi-step tasks with minimal input, starts pushing past developer labs into workplace tools. 2 3
This is significant as the contest is moving past just model benchmarks. According to the commission, Beijing’s push to weave AI into manufacturing, logistics, and robotics is churning out practical data that sharpens these models. The report also flagged “embodied AI” — essentially, AI running inside robots and autonomous tech — saying China’s advantage could accelerate if these systems keep rolling out. 6
The commission pointed to estimates showing roughly 80% of U.S. AI startups are now relying on Chinese open-source models. DeepSeek’s R1 managed to overtake ChatGPT on Apple’s U.S. App Store last year, it noted. Over on HuggingFace, Alibaba’s Qwen models have surpassed Meta’s Llama lineup in total downloads. 7
Still, some Western firms are giving it a shot. Siemens CEO Roland Busch says he sees “no disadvantages” in tapping Chinese open-source AI for the German company’s industrial models. According to Busch, a number of Siemens developers actually favor these tools over closed U.S. options—they’re cheaper on token costs and easier to customize. (A token, for the record, is the tiniest data chunk an AI model handles.) 5 4
DeepSeek’s budget-friendly models set off a global tech stock slide last year. Just last week, Alibaba’s U.S. shares dropped over 6% post-earnings, despite a 36% jump in cloud revenue. Xiaomi, listed in Hong Kong, surged as much as 5.8% after revealing that the mysterious model climbing OpenRouter’s leaderboard was its own. 8 9
The open-source surge comes with its own set of hazards. Chinese authorities have flagged security issues linked to OpenClaw-like agents, while Alibaba International Vice President Kuo Zhang has called out the “greatest risk” in deploying general-purpose models for sensitive business functions—that’s why their latest tool demands explicit user approval for payments or private file access. Siemens, for its part, notes some Chinese partners still won’t hand over the factory data needed to properly fine-tune industrial AI. 2 3 10
American companies still control key resources, from capital to chips. According to the commission, OpenAI and Anthropic have poured billions into maintaining their lead. And just last week, Nvidia confirmed it restarted making its H200 chip for China, now tweaked to fit U.S. export rules after Washington granted the necessary licenses. 11 12
Yet, according to commission vice chair Michael Kuiken, the U.S. is already running into “a bit of a deployment gap” when it comes to embodied AI compared to China. “That’s something that over time compounds itself,” Kuiken told Reuters. “We’re starting to see that compounding now.” 13
