Moonshot AI Releases Kimi K3.1 Pro with Cyber Restrictions
Moonshot AI introduced Kimi K3.1 Pro, a 3-trillion-parameter open-weight model with restricted cyber-offensive capabilities to align with proposed Chinese export controls.
Beijing-based startup Moonshot AI released Kimi K3.1 Pro, the first major upgrade to its flagship AI model. The 3-trillion-parameter-class open-weight model is designed to improve performance in coding, financial research, and 3D modeling for game design, targeting areas of current U.S. developer focus.
To prevent the model from extending the absolute frontier of cyber capability, the company removed cyber-related data from its training corpus and reinforcement-learning environments. These restrictions reduced the model's ability to find and exploit software vulnerabilities from 90.5% to 81.0% in internal benchmarks. Moonshot AI plans to publish the model weights by November 4, alongside a faster version called K3.1 Pro Highspeed.
This release coincides with proposals from the Government of China to implement export controls. The proposed regulations would ban AI-driven malware-generation technology and restrict AI-assisted vulnerability-discovery tools to protect government systems and critical infrastructure.
