Kimi K3 Punctures the AI Moat. Regulation Is the Next Fight.

The Chinese startup Moonshot released an AI model called Kimi K3 that matches American rivals on benchmarks at a fraction of the cost. It was built with open-source software and released as an open-weight model, meaning anyone can run it, modify it, or build on top of it without paying a license fee.

The competitive moat that U.S. AI companies have relied on – proprietary architectures trained on billions of dollars of compute – suddenly looks thinner. If a small Chinese shop can catch up in weeks on a shoestring budget, the advantage of scale erodes. The Mint editorial board, in a piece published this week, called the development a shake-up for "Big AI" and warned that the lack of a global safety framework makes the situation more dangerous.

Washington has imposed export controls on advanced AI chips and software, aiming to keep cutting-edge capabilities out of Chinese hands. But the policy shifts with the administration. Beijing has not restricted its own AI developers from releasing open-weight models globally. The result is an asymmetric market: U.S. companies face export limits that raise their costs and limit their addressable market, while Chinese models flow freely to any user anywhere.

The Mint board proposed a multilateral AI regulator modeled on the International Atomic Energy Agency, with India's 2026 presidency of Brics as the forum to start talks. The European Union's AI Act could serve as a template, the editorial said. That is a long shot – IAEA-style treaties take years to negotiate and longer to enforce – but the underlying argument is that no single country can police AI safety alone.