Chinese AI Firms Voluntarily Strip Hacking Capabilities from Open-Source Models

Chinese AI firms including Tencent (Hy4) and Moonshot AI (Kimi K3.1 Pro) are voluntarily removing offensive hacking capabilities from open-weight models while keeping defensive ones. Observers question whether this approach can remain sustainable amid intensifying competition and advancing AI capabilities.

2026-11-16 ~ 2026-11-16 · 3 related posts