It’s just over one A week ago I went to a major artificial intelligence Conference in Zhongguancun – Beijing’s buzzing high tech district
It was packed with fascinating sessions touching on everything from recursive self-improvement—the idea that models can tweak their own code and advance indefinitely—to humanoid robots. The event featured legends from the world of computing including Whitfield Barto and Rich Sutton who jointly won a Turing Award for their pioneering work in reinforcement learning.
One thing I took away from my trip was that the US and China need to put aside their intense AI rivalry.
Frontier AI’s cybersecurity, systemic risks and ever-more capable agentic model could cause chaos in the near future unless world superpowers work together. “AI is a global technology with global benefits, global harms, and a consistent tendency for new capabilities to eventually proliferate,” Stephen CasperAfterward, a computer-scientist at MIT spoke via video at the Conference.
Up until now, US policymakers have largely seen China’s AI developments as a threat to national security and the economy. Washington has placed strict restrictions on the chips and equipment used to make them in order to prevent China’s AI development. The US government has recently imposed strict restrictions on chips and chipmaking equipment to stymie the development of powerful AI in China. ordered The most powerful Anthropic models Mythos, Fable 5 and Fable 6 are not available to foreigners. national security concerns. Anthropic responded by rescinding access to everyone. WIRED revealed previously that one company was of special concern. South Korean telecom giant with alleged ties to China.
The Beijing Academy of Artificial Intelligence conference reinforced that the US as well as China will lose out if AI development is done too fast and irresponsibly. With AI becoming more powerful, agentic and integrated with daily life, there will be an increase in the risk that it can be misused to launch cyberattacks, or cause catastrophic outcomes. Cooperation between the two AI powerhouses is crucial because they are the ones responsible for most of the advanced models.
Casper pointed research Showing that international cooperation on AI threats outweighs any risks to national security that may come with working together. He compared the situation today to the time when the USSR and US were forced to collaborate on the dangers of nuclear weapons, even though they tried to surpass each other in stockpiling.
“One thing that almost everyone in AI can agree on right now is that AI doesn’t need a Chernobyl moment,” Casper says.
The cyber threats raised by AI are universal. There are also new types of vulnerabilities. AI-generated codeAgentic tools enable novel methods of attack on systems, as well as automated ways to carry out attacks. social engineering attacks.
Lin Yun is a Shanghai Jia Tong University professor who spoke to me after another session. does excellent work AI and computer safety. Yun said that he believes hackers will gain the upper hand in the short term. However, new countermeasures and novel AI applications should shift the balance over time.
Yun argued that, even though international cooperation was complicated by the competition, they should be prioritized. “If different countries understand the risks in similar ways, it becomes easier to develop shared safety principles and technical standards,” I heard him tell me. “The key is to find areas where sharing can reduce systemic risk without exposing sensitive operational details.”
The most pressing issue for both countries is to find a balance between openness and risk. Chinese-made open-weighted models, which are popularized in the US, have proven to be crucial for innovation and research. As these models progress, it becomes more difficult to make sure they do not help hackers find security flaws and cannot be used as cyber-weapons.

