OPEN AI MODELS CLOSING GAP WITH CLOSED SYSTEMS

The AI Security Institute, which has monitored frontier model cyber capabilities since 2023, found that open source models are accelerating toward parity with closed commercial systems. The narrowing gap reflects rapid advancement in publicly available AI development. Open weight models—which allow researchers and developers to access underlying weights and parameters—are catching up faster than previously observed throughout 2025. The institute's evaluation framework tracks cyber capabilities across frontier models, measuring performance on security-related tasks and vulnerabilities. The compression of the competitive gap suggests open source development is moving at an accelerated pace. This trend has implications for AI security strategy. As open models gain capabilities faster, organizations must adapt approaches to monitoring and controlling model deployment across both open and closed ecosystems.