Former Google CEO Eric Schmidt: U.S. and China Should Cooperate to Ensure Human Control Over AI Tools

On July 26, at the opening of the 2025 World Artificial Intelligence Conference (WAIC), former Google CEO Eric Schmidt stated during a dialogue with Harry Shum, Chairman of the Hong Kong University of Science and Technology Council, that China has made remarkable achievements in AI over the past two years and emphasized the need for U.S.-China collaboration.

Schmidt noted that the world has witnessed powerful AI models developed by Chinese companies like DeepSeek, Kimi, Minimax, Baidu, and ByteDance in recent years, with these technologies reaching world-leading levels. "I want to congratulate China's peers and enterprises on these accomplishments," he said.

Regarding the current state of AI development, Schmidt pointed out a key difference between the U.S. and China: in China, leading AI models are predominantly open-source, while in the U.S., top models are closed-source and do not release their weights.

He argued that the U.S. and China should collaborate to maintain global stability and ensure human control over AI tools. "China achieved modernization through reform and opening-up, attaining tremendous success today. Take Shanghai as an example—its product quality and robotics technology are truly impressive."

Schmidt described himself as an optimist, believing that the U.S. and China can gradually rebuild trust, as they have done historically and can do again now.

He also highlighted that companies are currently formulating strategic plans, with OpenAI and Gemini making progress while DeepSeek and others excel. Over the next two years, he predicted a surge in AI agents that will profoundly transform workflows in businesses and governments—an exciting prospect.

If development proceeds smoothly, Schmidt foresees advancements in deeper reasoning and more sophisticated algorithms. While the boundaries of technological progress remain unclear, he stressed the importance of consensus-building among China, the U.S., and other nations through dialogue at critical junctures. For instance, if AI autonomously seeks weapons, self-replicates, or learns without authorization, the response shouldn’t be a simple shutdown but guided discussions to prevent失控 (loss of control).

Schmidt concluded by expressing optimism for the next five years: "We will witness positive societal transformations, and the world will become a much better place in many ways."