HMM: China AI labs face growing open-source dilemma.
It also underscores a new reality in Silicon Valley and beyond: startups and more established companies alike are increasingly eschewing pricey proprietary models from OpenAI and others for free or low-cost Chinese versions which are rapidly closing the performance gap despite Washington’s export curbs and sanctions. Rankings on OpenRouter, a platform that lets developers access and use different models, currently show 7 of the 10 most popular offerings are Chinese; companies from Airbnb (ABNB.O), opens new tab to German industrial giant Siemens (SIEGn.DE), opens new tab have made no secret about using models from the People’s Republic.
Chinese tools, however, now have a target on their back. On Monday, the U.S.-China Economic and Security Review Commission published a report, opens new tab warning that China’s open-source successes “reflect a more fundamental challenge to U.S. AI supremacy”. And last month, OpenAI and Anthropic separately accused Chinese AI labs like DeepSeek of improperly using U.S. models to improve their own; both have called for tougher restrictions on U.S. tech.
There is a simple reason for Chinese companies to pivot away from open-source: Alibaba, Tencent, and rivals are under increasing pressure to show they can monetise their AI models and applications.
Well, “Dump, dominate, and monetize” has been China’s business model for three decades.
