I Tested Kimi K3 on a Real Astro Codebase: Strong Cross-File Analysis, Unsafe First Fix

Kimi K3 trails frontier US models by a wide margin on cyber exploits, and distillation may explain why

Kimi K3 trails U.S. frontier models 32.2% vs 76.2% on cyber exploits (ExploitBench), achieving no ACE on 41 Chrome vulnerabilities. Distillation from Claude without offensive cyber data (filtered by Anthropic's safety classifiers) explains the gap, revealing U.S. safeguards as a lasting competitive moat.