China calls for global effort to guide AI development

SHANGHAI — The development and governance of artificial intelligence should be a global effort, Chinese President Xi Jinping said Friday, while reiterating China’s objections to what he called “overexpansion” of national security concerns.

Speaking at a conference in Shanghai, Xi said AI should not be dominated by a single nation. U.S. restrictions have prevented China from accessing some of the world’s most advanced technologies, boosting China’s efforts to develop its own know-how and intensifying rivalry between the world’s two largest economies.

“The development of artificial intelligence should not be the work of a single country, but rather a symphony of global cooperation,” Xi said at the opening of the annual China World Artificial Intelligence Conference in Shanghai. Other participants included the leaders of Kazakhstan, Cambodia and Thailand, as well as UN Secretary-General António Guterres.

“We should together oppose the practice of overextending the concept of national security in the field of artificial intelligence and placing its own security above that of other countries,” he said, reiterating a long-standing Chinese complaint.

China will expand AI cooperation with the Association of Southeast Asian Nations, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organization and other BRICS countries, Xi said. He promised to give 30 countries access to a Chinese-developed AI weather tool that provides early warning systems.

Over the next five years, Xi said China will provide 5,000 AI training opportunities to developing countries.

Closer partnerships can help prevent “historic injustice in AI,” he said.

Ahead of the conference, 29 countries, including Pakistan, Russia and Kazakhstan, signed an agreement with China to establish a Global Artificial Intelligence Cooperation Organization. State media have described it as an intergovernmental organization headquartered in Shanghai that promotes global AI governance.

The new AI cooperation organization can be seen as China’s response to the US Pax Silica initiative, said George Chen, partner and chairman of the digital practice at Washington-based consultancy The Asia Group.

The Pax Silica framework, launched late last year, focuses on strengthening collaboration with U.S. allies and partners on AI-related supply chains. Signatories include Japan, the United Kingdom, Australia, the Philippines, Israel and India.

Following US President Donald Trump’s visit to Beijing to meet Xi in mid-May, China and the US also agreed to hold a dialogue on AI development and governance.

Chen, who was present at the conference in Shanghai, also said that Xi’s speech can be seen as a signal that China can be a reliable partner for the developing world, or countries of the “South.” “China will not let America have a monopoly on AI technology.”

More than 1,100 companies and 1,400 guests will attend this year’s annual AI conference, Chinese state media said.

LATEST MOONSHOT MODEL

Before Xi delivered his full speech, another powerful new artificial intelligence model from China stunned the U.S. tech industry Friday, the latest sign that Chinese startups publicly releasing their “open source” AI technology are making California’s AI titans sweat.

The latest Kimi K3 model from Beijing-based startup Moonshot appears to be catching up with the best versions of Anthropic’s Claude and OpenAI’s ChatGPT.

“This may be the biggest release of the year,” and marks the moment when open-source Chinese models overtake closed U.S. models, said Anastasios Angelopoulos, co-founder and CEO of Arena, a platform for evaluating AI systems.

Kimi K3 tops the Arena rankings in what it calls “front-end coding capability,” which is a measure of the performance of a large AI language model. “More results are coming that should continue to show he is leading the pack,” Angelopoulos said on social media.

K3 follows another major release of an AI model last month from Chinese startup Zhipu, or Z.ai. Its new flagship model, the GLM-5.2, is already widely used by software developers around the world, who say it can do work almost as well as the best American models at a lower price.

The hype around the new Chinese model resembles the panic that shook the market that followed the release of a model from Chinese startup DeepSeek in early 2025, although not everyone finds it justified. The response to K3 is an “overreaction and similar to DeepSeek last year,” tech analyst Patrick Moorhead said on social media. He said this could be beneficial for parts of the broader AI industry, but would pose a revenue problem for Anthropic and OpenAI.

During the conference that runs through Monday, tech giant Huawei also showed off a new AI computing system called Atlas 950 SuperPoD, a sign that China is increasingly amassing the domestic hardware it needs despite U.S. restrictions on imports from chipmakers like Nvidia.

Moonshot didn’t specify what hardware it used to build the K3, but the startup partners with Huawei.

The price to use K3 is the highest ever for a Chinese AI model, but it’s still half the price of OpenAI’s high-performing GPT-5.6 Sol model, according to a report released Friday by Bank of America analysts.

Information in this article was provided by Han Guan Ng, Chan Ho-him, Ken Moritsugu and Matt O’Brien of The Associated Press.