Japanese Media: China's AI Models Trigger Panic-Driven Debate in the U.S. Industry

An article published by Business Insider (Japanese edition) on July 22 noted that the stunning debut of China’s AI model Kimi K3 has drawn intense attention from Silicon Valley. Divisions among leading figures in American artificial intelligence are deepening over the question of "what is the best strategy to win the AI development race." Some leaders at U.S. AI companies have begun embracing inspiration from China—specifically, the proactive adoption of open-weight AI models.

Every time a Chinese company announces a new AI model, it triggers panic in the United States. On July 16 (Japan time, July 17), the announcement of "Kimi K3" by China’s Moonshot AI sent shockwaves through the American tech industry. The model achieved overwhelming results in major benchmark tests, with most observers stating its performance rivals that of America’s leading cutting-edge models—and at an extremely low cost.

As a result, the U.S. tech community has launched a fierce debate over “open versus closed” approaches.

The emergence of the new "Kimi K3" model has sparked fresh fears that "China is catching up to the U.S., vying for dominance in the AI market." At the same time, criticism has erupted, accusing China of leveraging achievements already established by companies like Anthropic, OpenAI, and Google to train its models. China actively adopts "open-source" models, publicly sharing training data and methodologies, as well as "open-weight" models—releasing only the final trained results—while the U.S. maintains a closed system where most models remain undisclosed. With open weights, developers can verify model content, independently customize, and deploy them.

OpenAI Executive’s Shocking Statement: Over the weekend, amid the "Kimi K3" panic, an OpenAI executive posted a lengthy comment about Kimi K3 on X, sparking intense controversy.

"Given the potential risks, I am personally surprised that the Chinese government continues to allow such high-performance models to be open-sourced," said Dean Ball, OpenAI’s strategic lead, in a post on X. Ball previously served as a senior AI advisor to President Trump. He argued that the open-weight strategy would ultimately lead to a complete "AI communism." He believes open models may progress slowly and hinder the evolution of AI because they "discourage capital investment in AI."

However, what truly ignited the debate was the following statement: "I predict that the Trump administration will eventually realize that using open-weight AI models from China can be deliberately weaponized to create mass panic." Ball suggested: "By intentionally manufacturing fear, uncertainty, and doubt (FUD) during the regulatory process, we can prevent other companies from adopting open models."

Later, Ball clarified that this was merely his personal prediction, not a recommendation. He stated he would only support open-sourcing once AI reaches a stage deemed "too dangerous"—adding that "reaching that point would be a sad day." The idea of deliberately creating regulatory obstacles to benefit U.S. AI labs sounds exactly like what Nobel laureate George Stigler described as "regulatory capture"—a situation where regulatory agencies design rules favoring the industries they oversee, based on advice from insiders within those industries.

Both Anthropic and OpenAI have consistently argued that their models are too powerful to be opened. Their logic is that once such tools are made public, anyone could use them for any purpose without comprehensive oversight, making them dangerous. By maintaining closed systems, developers retain strong control over safety, access rights, pricing, and other aspects.

Both companies have warned that unrestricted weight models originating from China pose threats to national security and their own business interests.

David Sacks, a venture capitalist who served as the first AI and crypto assets advisor in the Trump administration and became co-chair of the Presidential Council of Advisors on Science and Technology (PCAST) in March 2026, responded to Ball’s post on X: Weaponizing regulation is "completely unacceptable." "We are now at a critical turning point in AI policy. In terms of profit from AI models, we are effectively in a duopoly. These major closed AI companies want to use government power to eliminate open-source competitors."

Original Source: toutiao.com/article/1871380776265799/

Disclaimer: This article represents the views of the author alone.