Kimi K3 moment should prompt some to rethink AI's future for common good

Whenever a Chinese artificial intelligence lab drops a new high-performance model, there are observers in the United States who lose their composure. After DeepSeek in 2025, it happened again last week, when Moonshot AI released Kimi K3 — a 2.8 trillion-parameter open weight model that scored an Elo rating of 1679 on the Frontend Code Arena leaderboard, surpassing both Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on this specific coding benchmark.

Executives at the US' leading closed model AI labs began floating ideas about how to stop the open source "menace" — not by building better models, but by enlisting Washington to create "regulatory risk" around Chinese open weight models.

Ironically, the country that invented a large part of the modern tech industry through open standards and competitive markets is now debating whether to use the power of the state to protect a handful of companies from competition.

The US' leading closed model labs — essentially a duopoly of OpenAI and Anthropic — have spent years arguing that their models are too dangerous to open source. Now, their defenders are suggesting that the US administration should manufacture fear, uncertainty and doubt around Chinese alternatives to keep them out of the market.

Dean Ball, OpenAI's new head of strategic futures, was refreshingly candid about this on X. He acknowledged that Kimi K3 is "a very good model" whose performance cannot be "explained away by distillation or anything like that". Then he urged the US administration to "create large amounts of regulatory risk around the use of open weight Chinese models" — not through outright bans, but through "soft law" that creates FUD (fear, uncertainty and doubt).

This is "regulatory capture" dressed up as "national security". As David Sacks, the venture capitalist who now cochairs the President's Council of Advisors on Science and Technology, put it: "The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition. They have laid their cards on the table."

China has achieved frontier-level AI performance at a fraction of the cost. Kimi K3's API pricing is $3 per million input tokens and $15 per million output — roughly one-third to one-fifth the cost of comparable US models. And Moonshot AI has announced it will release the full model weights soon, meaning anyone with sufficient compute can run it on their own infrastructure without paying per-token fees. That would be unimaginable for the US companies that charge high fees on tokens.

But the real story here is about two different philosophies of AI development colliding. The US closed model is built on concentration: a handful of corporations and the US government standing at the apex of the AI pyramid, nurturing and reaping the benefits of everything beneath them. It is an architecture where a small number of gatekeepers decide who gets access to frontier capabilities and at what price. The Chinese open weight approach is more inclusive.

These two paradigms differ on many fronts, from profitability and costs to data acquisition, application scope, iteration efficiency and the boundaries between public interest and private gain. The over 40 participants of the Global Partnership on Artificial Intelligence and the founding members of the World Artificial Intelligence Cooperation Organization — established in Shanghai last week — should not devolve into two camps. The world does not need a Cold War replicated in the AI era.

The deeper question is whether the US can adapt to a world in which it no longer holds a monopoly on cutting-edge AI. The Kimi K3 moment should be the beginning of a reckoning. Will Silicon Valley respond with innovation or with "regulatory capture"? If AI is to serve humanity, the world must find the wisdom to resist the zero-sum temptations of shortsightedness and anxiety.