Ensuring national security is a top priority for governments. To accomplish this task, artificial intelligence (AI) has become a critical tool. The algorithmic core of AI systems typically consists of models trained using machine-learning techniques. These models can be subject to varying degrees of openness. Computer science commonly distinguishes between open and closed source. In the context of AI models, another category has emerged between these two poles, known as “open weight.” The goal of this review protocol is to outline a systematic literature review on the impact of open-source and open-weight AI models on national security. This document defines the review questions, inclusion criteria, a search strategy, and methods for data collection and analysis.
Koch et al. (Thu,) studied this question.