AI Models Know What Women, Peace and Security Is. They Just Don't Bring It Up (blog.cip.org)

The Collective Intelligence Project (CIP) has launched the Weval platform to assess AI models' adherence to Women, Peace and Security (WPS) principles, highlighting the "WPS Competence Gap," where AI tools fail to account for the experiences of over half the global population, particularly affecting peacekeeping decisions. Recent evaluations demonstrated that leading large language models (LLMs) like Gemini 3.1 Pro and GPT-5.6 Sol routinely omit gender considerations, reducing the efficacy of peacekeeping strategies and inadvertently supporting systemic biases prevalent in military and security operations. This gap is critical as it could undermine decades of progress established by the UN's 2000 Resolution 1325, which emphasizes women's roles in peace and conflict resolution. The study underscores an urgent need for AI systems to incorporate gender perspectives into their analyses to avoid reinforcing existing inequalities. By setting WPS benchmarks, CIP aims to encourage AI vendors to enhance their models, ensuring that tools used in decision-making reflect a comprehensive understanding of social dynamics, ultimately leading to more effective peacekeeping initiatives that are inclusive of all voices.