PHOENIX (AZFamily) \u2014 More than 1,300 employees and executives at OpenAI, Anthropic, Google, Meta and other AI companies signed a letter last week asking the U.S. government to help build tools to deliberately slow AI development.<\/p>\n
Chinese AI companies spent the same week accelerating.<\/p>\n
On July 27, the Beijing startup Moonshot AI<\/a> released the full weights of Kimi K3<\/a>, a 2.8-trillion-parameter model anyone can download and run for free (although it requires expensive, commercial-grade computing power). Independent benchmarks put it within striking distance of the best American systems.<\/p>\n Arizona has a direct stake: some of the advanced Nvidia chips<\/a> at the center of the dispute are made at TSMC\u2019s plant in north Phoenix<\/a>.<\/p>\n Lily McElwee, president and CEO of the Phoenix Committee on Foreign Relations<\/a> and an expert on U.S.-China relations, said the release put nearly every unresolved question in the rivalry on the table at once.<\/p>\n Open weights as foreign policy<\/p>\n McElwee said Beijing\u2019s embrace of open-weight models \u2014 systems whose parameters are published, so anyone can download, modify and run them instead of paying for access \u2014 fits a much older strategy.<\/p>\n \u201cChina has always framed itself as a provider of more freely accessible, low-cost technology to the rest of the world while the United States provides these proprietary systems,\u201d she said. <\/p>\n The pitch is aimed at developing countries and the Global South, she said, and is a way to frame American firms as primarily motivated by profits.<\/p>\n Asked whether the open-weight push resembles China\u2019s dumping of subsidized steel to drive out competitors, McElwee said the comparison holds.<\/p>\n \u201cI\u2019m not saying that Beijing is acting in good faith,\u201d she said. \u201cI\u2019m saying it\u2019s a marketing ploy.\u201d<\/p>\n She pointed to Huawei and 5G as the closest precedent.<\/p>\n Huawei grew into the world\u2019s largest supplier of telecom network equipment, in part by undercutting Western rivals on price with the help of state subsidies and cheap state financing for developing-world customers.<\/p>\n The U.S. blacklisted the company in 2019 and pressed allies to keep its gear out of their 5G networks. Much of Europe, Africa and Latin America built 5G networks with at least some of it anyway.<\/p>\n The security case<\/p>\n A Booz Allen Hamilton study published in June found<\/a> that three of four Chinese models tested produced significantly more vulnerable code when the user was identified as working for the U.S. government.<\/p>\n The concerns run in an obvious direction, McElwee said.<\/p>\n \u201cWashington is rightly concerned about data vulnerabilities when it comes to U.S. customers using open weight Chinese models, because the output of those models is generating data on U.S. citizens,\u201d she said.<\/p>\n McElwee noted the U.S. bans Chinese electric vehicles and other Chinese technology products. \u201cBut if we\u2019re not banning AI systems, which are increasingly used to train those products and access consumer data in the United States, then we have a security vulnerability.\u201d<\/p>\n Kimi K3 has drawn two further accusations. Michael Kratsios, director of the White House Office of Science and Technology Policy<\/a>, claimed that Moonshot trained K3 on outputs from Anthropic\u2019s Claude Fable 5 and used restricted Nvidia chips routed through Thailand. Moonshot has not responded publicly.<\/p>\n Where export controls stand<\/p>\n The Biden administration imposed the first significant restrictions on advanced AI chip exports to China. It was a core grievance for Beijing, McElwee said, because chips are its choke point in the race.<\/p>\n The Trump administration eased some controls in January, moving Nvidia\u2019s H200 from near-automatic denial to case-by-case review. The most advanced Blackwell processors remain off limits.<\/p>\n The practical picture has not changed much, she said.<\/p>\n \u201cThere\u2019s not a blanket restriction on it anymore. It\u2019s a case-by-case basis,\u201d she said of the H200. \u201cBut to my knowledge, there have been no cases that are approved.\u201d<\/p>\n China is getting the chips anyway, she said, through partners in Southeast Asia, stockpiling and recruiting the scientists who design them. She said that is where enforcement should focus.<\/p>\n Washington has also taken action on imports, with the FCC banning new imports of foreign-made humanoid robots<\/a>, robot dogs and power inverters last week on national security grounds. The administration has been weighing restrictions on Chinese open-weight models as well, although several outlets reported Tuesday that the White House will exempt open-weight models from its review regime.<\/p>\n McElwee is skeptical that restrictions on open-weight models would work, given that the models are freely downloadable by design. She argued that controls should target the tangible advantage the U.S. actually holds: the chips.<\/p>\n \u201cLet\u2019s take advantage of our strong points in this AI race, which are advanced Nvidia chips, which are produced right here in Arizona,\u201d she said.<\/p>\n Watch the full interview on Generation AI, streaming on AZ Family+ and at azfamily.com\/GenAI<\/a>. New episodes post Wednesdays at 7 p.m.<\/p>\n See a spelling or grammatical error in our story? Please click here to report it<\/a>.<\/p>\n