China’s leading AI companies are beginning to restrict the cyberattack capabilities of advanced open-weight models while preserving their ability to defend against cyber threats. This month, Tencent said it had removed the model Hy4’s hacking capabilities, following a similar measure taken by Moonshot AI for Kimi K3.1 Pro. Drawing on several AI hacking incidents involving U.S. and Chinese companies, the article examines whether these adjustments can reduce risks without undermining the models’ competitiveness.
Source:scmp.com · Read the original article
Leave a Reply