Can Chinese AI companies curb cyber risks from open-weight models without losing their edge?
China’s leading artificial intelligence model companies are beginning to curb the cybersecurity capabilities of their most advanced open-weight models, following a string of high-profile AI hacking incidents involving both US and Chinese companies. Earlier this month, Chinese technology giant Tencent Holdings said it had stripped its latest Hy4 model of its cyber-hacking capabilities while…
Leading Chinese artificial intelligence companies are beginning to limit the cybersecurity capabilities of their advanced open-weight models following several high-profile hacking incidents. SCMP Tech reports that Tencent Holdings recently removed cyber-hacking capabilities from its latest Hy4 model.
Despite U.S. export controls on high-end AI accelerator chips, research from UnemploymentLabs suggests that Chinese firms are still accessing Nvidia Blackwell hardware to train models. According to The Register and Tom's Hardware, forensic evidence indicates that companies including DeepSeek, Moonshot AI, MiniMax, and Ant Group have used these chips to develop models such as DeepSeek V4 and Kimi K3.1.
While the research identifies hardware signatures within the models, UnemploymentLabs notes it does not prove export controls were broken or reveal how the chips were accessed, as renting hardware abroad is not currently illegal.
AI entrepreneur Kai-Fu Lee told Fortune Daily that U.S. restrictions have pushed Chinese developers to build competitive models using fewer resources. Lee estimates that these Chinese open-weight models are currently only six to 12 months behind those produced by major American firms like OpenAI and Anthropic.
Brief written by urgent.news from SCMP Tech, Tom's Hardware, Fortune, The Register, The Register Science — 5 reports on this story. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
