(July 17): President Xi Jinping has capitalized on the rise of China’s artificial intelligence (AI) model to stake his claim to shaping global rules for technology, even as its growing power has raised security alarms in Washington and Beijing alike.
Xi emphasized safety and equality in his debut at the World AI Conference in Shanghai on Friday, and called for broader international cooperation at the event, which previously included Elon Musk and Jack Ma. He was speaking to a crowd of technology companies and government leaders as U.S. companies’ share of AI use in the popular OpenRouter market nears a record 60%, and the Chinese model attracts companies around the world.
“We must engage in broad international cooperation to support countries in the Global South in building capacity to close the AI and digital divide,” he said, calling for efforts to avoid creating “historic injustices in AI.”
His remarks echoed the drumbeat of articles in state media that framed China’s openness as an antidote to a walled world. Representative spokesperson of the Communist Party People’s Daily Earlier this week, he warned against an “Iron Curtain of AI” and contrasted the “oil mentality” of storing data and computing power with the “water mentality” of treating AI as a public good for all, without naming countries.
The Chinese government is now considering competing with the United States, an AI leader, for influence through a new group of about 30 countries called the Global AI Cooperation Organization. Xi defended the bloc in his keynote address, which was proposed last year and established on Thursday, pledging to align global AI rules and technology standards to “bring this frontier technology to better benefit humanity.”
Researchers in the country openly say the organization has ambitions to provide China with a platform to shake up global rules and standard-setting around AI.
“China is not only the world’s largest open source application market, but also a major contributor. However, its voice in shaping international rules still lags behind China’s power,” said Gu Lingyu, an AI researcher at Peking University.
“Effectively telling the story of China’s open source practices can enhance its ability to shape global open source governance,” Gu said at a roundtable discussion published by the journal of the Supreme People’s Court.
But behind the rhetoric of public interest, the Chinese government faces the same dilemma as the US government. As models become more capable, the question is how to balance mass adoption with national security.
Chinese authorities recently held talks with companies including Alibaba Group Holding Ltd., the developer of the popular Qwen model, about ways to reduce security risks posed by the increasingly powerful models, according to people familiar with the matter. Negotiations are in the early stages and there are no plans for enforcement, but restricting access to top models from abroad was an option, the people said.
Reuters It was previously reported that the Chinese government was considering restricting overseas access. Alibaba and the Department of Commerce did not respond to requests for comment.
In his speech, Xi emphasized the need to curb security risks posed by rapidly advancing technology and cautioned against overextending the concept of national security.
The warning echoes that of Washington. The United States last month temporarily banned foreigners from accessing Anthropic PBC’s Mythos and Fable models on national security grounds, as that class of systems can exploit well-hidden software flaws, sometimes without human oversight.
“The Chinese model may achieve Mythos-level cyber capabilities in the coming months,” said Saif Khan, a distinguished technology researcher at the Institute for Progress. “This is a real security risk and one that many parts of the Chinese government will be concerned about.”
Oversight from both governments has put pressure on companies building openly available models, raising concerns that a crackdown by either capital could destroy global research networks and the proliferation of start-ups.
Flo Crivello, founder of AI assistant startup Lindy AI, said China’s production shutdown “would have a pretty big impact.” His company moved from Anthropic’s models to DeepSeek last month and is now saving 90% in the cost of running inference, or AI.
“This effectively cancels out these savings until the open source scene catches up in the US,” he said. But he doesn’t see losing access as “the end of the world,” predicting it will take three to six months for companies like Alphabet Inc.’s Google and Meta Platforms to release free-weight models to rival Chinese products.
The introduction of Chinese AI by American companies is itself under fire. In April, lawmakers launched an investigation into Airbnb Inc. and Cursor developer Anysphere Inc. over their use of Chinese models. Airbnb CEO Brian Chesky defended the choice to use Qwen for its customer service chatbot, saying no user data would be sent to foreign companies.
U.S. AI companies and the White House have also accused China of inappropriately extracting features from the U.S. Frontier model, an accusation that Beijing denies. Shyam Sankar, Palantir Technologies’ chief technology officer, this week called China’s open model an economic threat and accused it of stealing American intellectual property.
None of this is likely to shake up Beijing’s strategy, which has turned openness into its main competitive advantage.
Chinese products first overtook American rivals in OpenRouter’s U.S. token traffic earlier this year, reaching 63% in the first week of July, and DeepSeek has become the most popular choice among U.S. companies in recent months.
This measurement does not capture direct use of models provided by Anthropic or OpenAI. But it shows how much the open-weight approach — a model where anyone can download and run it for free — has caught on with Chinese developers, with U.S. companies using the platform at a rate of less than 10% a year ago.
“Open source AI is a very important part of China’s AI strategy,” said Christy Rourke, a fellow at MATS Research who studies China’s AI governance. “This is not only an important part of China’s domestic penetration strategy, but also a critical catch-up strategy where everyone builds on top of each other.”
Rather than giving up that advantage, Roque hopes regulators will tweak their oversight, including reserving the toughest regulations for frontier companies.
Chinese industry leaders also advocate openness. Zhipu founder Tang Jie wrote in an internal memo: bloomberg news He argued that frontier AI should remain broadly accessible rather than controlled by a few, but stressed that safety is now a “fundamental prerequisite” for any technology powerful enough to change civilization.
Meanwhile, that confidence is spreading to the public eye. In a recent exchange on X, Elon Musk predicted that China could rival Anthropic’s top-tier Fable 5 in “probably the first quarter.”
Mr Tan replied: “It won’t take that long.”
Uploaded by Chong Xia Lane
