SerpApi.Org
Donald Trump faces a mounting battle with China after an AI breakthrough in the country threatened to upend US dominance of the technology.
The unveiling of Kimi K3, an open-source AI model developed by Chinese company Moonshot, has rattled stock markets and Western tech executives amid fears that it will challenge business models and US tech supremacy.
The model is as powerful as the most advanced AI systems launched by US leaders OpenAI and Anthropic, while being cheaper to run. Moonshot also lets people freely download and modify the code underpinning the model.
Moonshot was forced to pause new subscriptions on Sunday after a stampede of demand to use its new model overwhelmed its servers.
Separately, Chinese tech giant Alibaba said it too was preparing to launch a new version of its open-source AI model Qwen that it claimed was “one of the most powerful models available” and only just behind Anthropic’s most advanced system.
The releases come as Xi Jinping, the Chinese president, seeks to promote Chinese AI across the developing world in a move that could embed the Beijing-controlled technology across nations.
Dean Ball, a senior executive at ChatGPT-developer OpenAI, wrote on X that China’s strategy of giving away its AI code would probably lead to “full AI communism, which is precisely what China proposes”.
Tech stocks around the world crashed last week and there are fears that the slump will continue on Monday after investors spent the weekend getting to grips with the implications of Kimi K3.
Mr Ball suggested that the Trump administration would look to limit the use of Chinese AI models as a result. Alibaba has been blacklisted by the White House because of national security concerns.
Kimi K3 is the world’s largest open AI model by parameter count, a measure of power, and analysis by Artificial Analysis and Arena.ai showed that it had a similar performance to the most advanced models developed by the US’s OpenAI and Anthropic.
Moonshot’s model is also cheaper to run, according to Artificial Analysis, at $0.95 (£0.70) for each intelligence index task compared to $1.04 for OpenAI’s most advanced model.
Kimi K3’s launch has triggered much soul searching in the US about why it was not built there. Yang Zhilin, Moonshot’s founder and chief executive, developed his skills as a PhD student at Carnegie Mellon University (CMU) in Pittsburgh, Pennsylvania, before returning to his native China.
Russ Salakhutdinov, a CMU professor who taught Mr Yang, claimed on X that his student had “many opportunities to stay in the US” including an offer from Apple but chose to return to his homeland.
Moonshot’s dramatic rise comes as Xi seeks to extend China’s AI influence globally. In a speech on Friday, he said Beijing would help developing countries scale up their AI capabilities.
It came just a day after 29 countries including Russia, Brazil and Pakistan signed up to the China-led World Artificial Intelligence Cooperation Organisation.
Prof Alan Woodward, a cyber security expert at the University of Surrey, said: “The Chinese seem to be letting their models out to the public.
“The US is trying to control their dissemination. It looks like a strategic decision that could see the Chinese dominate in the long term, at which point they could well start to be more controlling.”
Richard Windsor, an independent technology analyst, said: “China’s strategy in AI looks like a rinse-and-repeat of its highly successful takeover of the solar panel industry and its strategy in the automotive industry, which is also beginning to bear fruit.”
A recent report by US company Harmonic Security estimated that one in 12 workers in the US and UK were already using some form of Chinese AI. Alastair Paterson, Harmonic’s chief executive, said: “All data submitted to these platforms should be considered property of the Chinese Communist Party given a total lack of transparency.”
Not only would such a giveaway extend Beijing’s influence, it would also threaten to undermine the business models of US tech giants such as OpenAI and Anthropic.
These companies are betting that subscription fees for their AI will justify the substantial sums spent on developing the technology and help them achieve a $1tn-plus valuation.
Ciaran Martin, the former head of the National Cyber Security Centre, part of GCHQ, said: “An open weights model – that means it’ll soon be available to anyone to use and adapt it which, combined with its lower cost, will make it very attractive worldwide outside the US’s direct sphere of influence.
“So it puts pressure on the US frontier labs’ business models.”
Pressure on the US AI giants would have far-reaching implications. The US economy has largely been powered by unprecedented levels of investment in AI infrastructure and technology. Any hint that this would not pay off would trigger a sharp stock market crash at the very least.
The availability of such a powerful open-source AI model has also set off alarm bells among cyber security experts.
In June, the US government temporarily pulled access to Claude Mythos, Anthropic’s latest AI model, over fears that it could be used to hack financial systems.
Mr Martin warned Kimi K3 “raises security challenges”. He said: “That’s not just because it’s Chinese but because it’s open.
“It means it’ll have similar cyber attack capabilities to Claude Mythos but you won’t be able to gatekeep them as has been done in that case.”
