UK AI Security Institute

Open-Weight Models Narrow the Cyber Gap with Proprietary AI; Kimi K3 Emerges; Hassabis Proposes AGI Regulation

The UK AI Security Institute finds that open-weight models like GLM-5.2 and DeepSeek V4-Pro now lag frontier closed models by only 4-7 months, down from 6-10 months. Chinese firm Moonshot AI releases Kimi K3, a 2.8 trillion parameter model matching frontier performance, with weights to be open-sourced. DeepMind's Demis Hassabis proposes a FINRA-like standards body for AGI oversight.

UK AI Security Institute Moonshot AI DeepMind