China’s Kimi K3 calls US safety curbs on top AI models into question

China’s Kimi K3 well behind US rivals in cyberattack ability, study shows

Kimi K3 achieved 32.2% on cyberattack capability tests versus 76.2% for top US models, UK-US joint government research shows. The performance gap indicates China's frontier AI significantly lags in cyber-capability, easing Western near-term risks to critical infrastructure defense.