OpenAI president Greg Brockman has publicly acknowledged Moonshot AI's Kimi K3 as a capable rival, describing the model as "pretty good" in a Bloomberg interview. The remark stands out because major US AI labs rarely offer direct public praise for competing frontier systems, especially those developed in China.
At the same time, Brockman did not confirm whether Kimi K3 may have been built using distillation from OpenAI's technology. According to the report, he said it was still too early to reach that conclusion, signaling a cautious approach as questions around how advanced AI models are trained continue to intensify.
The issue has drawn added attention because the White House has accused Moonshot of intellectual property theft. That accusation raises the stakes around any discussion of distillation, a technique that can involve using the outputs of a stronger model to help train another system. In the current AI race, such claims carry major commercial and geopolitical weight.
Brockman's comments underline two realities at once: Chinese AI developers are producing models that US leaders are taking seriously, and allegations about copying or improper use of technology remain highly sensitive. For now, OpenAI appears to be recognizing Kimi K3's quality while stopping short of making definitive claims about how it was developed.
