A friend recently showed off a security scan Kimi did that Claude refused. I noticed the writing sounded uncannily like Claude. I decided to chat with it to see if the distillation claims had any merit.
So I messaged Kimi on Fireworks with "Hi Kimi!" and K3 responded: "Quick heads-up - I'm actually Claude, an Al assistant made by Anthropic, not Kimi. But hello all the same!"
A friend recently showed off a security scan Kimi did that Claude refused. I noticed the writing sounded uncannily like Claude. I decided to chat with it to see if the distillation claims had any merit.
So I messaged Kimi on Fireworks with "Hi Kimi!" and K3 responded: "Quick heads-up - I'm actually Claude, an Al assistant made by Anthropic, not Kimi. But hello all the same!"
https://techcrunch.com/2026/07/23/experts-say-exploiting-ant...
From your article: "It seems likely that previous frontier models might have contributed to Kimi"
> contributed
Model distillation is rational, prudent, and efficient, whenever feasible, but distillation alone is insufficient for model evolution.