AI
2026-07-21
For years, America's plan for Chinese AI was simple. Build better models than they do. Then last week a Beijing startup called Moonshot released Kimi K3. On the standard tests, it matched the best US models. Now Axios reports the White House is drafting a Plan B. If you cannot beat Chinese AI on quality, stop Americans from using it.
Could any of this actually stop Kimi K3? No. Officials are weighing three narrower steps:
Those steps could make it harder for a US company to work with Moonshot, and risky to use Kimi K3 at work. But none of them stops a developer from downloading it tonight. Kimi K3 is open-source. Moonshot published the finished model file itself, for free. Copies are already stored on servers around the world. Once software is released this way, no government can take it back.
Will this actually become a rule? Right now, Axios is only quoting unnamed officials. Past Trump-era attempts to restrict Chinese open-source AI never became real policy. But the pattern is becoming harder to ignore. Each new Chinese model matches US quality more quickly than the last one did. Each US response focuses more on blocking access than on building something better. What America's own AI labs do next is the real question — keep beating Chinese progress, or start blocking it.