UN Women Warns AI Perpetuates Gender Bias, Reinforcing Harmful Stereotypes

The United Nations Women organization has released a critical report highlighting how artificial intelligence, especially advanced large language models (LLMs), is reinforcing and exacerbating existing gender biases. The report, published on June 22, 2026, underscores concerns that as AI becomes more integrated into daily life, it risks embedding and amplifying harmful stereotypes that disadvantage women globally. A key finding from a study cited by UN Women revealed that a significant percentage of AI systems exhibit gender bias. Out of 133 AI systems analyzed, 44 percent demonstrated gender bias, and 26 percent displayed both gender and racial bias. This pervasive issue is evident in how LLMs process and generate language. These models frequently associate women with roles related to home, family, and childcare, while consistently linking men to business, leadership, and career achievements. In some instances, when prompted to complete sentences based on a person's gender, approximately one in five responses from LLMs were found to be sexist or misogynistic, even portraying women as sexual objects or property. UN Women emphasizes that these outcomes are not accidental errors or glitches but rather a predictable consequence of AI systems being trained on vast datasets that reflect decades of unequal representation of women and men in society. Jayathma Wickramanayake, UN Women Lead on Digital Technologies, explained that AI models "pull bias from decades of text written by people, about people, in a world where women were filed under home and family, and men were filed under business and career." This highlights a critical policy gap, as only a small fraction of countries have incorporated gender considerations into their national AI strategies. Beyond perpetuating stereotypes, the report also addresses the intensification of online violence against women and girls facilitated by AI. Data from UN Women indicates that nearly a quarter of surveyed women human rights defenders, activists, and journalists have experienced AI-assisted online violence. This includes the non-consensual sharing of personal images and the proliferation of deepfakes. As AI-generated content becomes more commonplace, the tools for harassment and manipulation are scaling, making it harder to detect and prevent such abuse. The organization is calling for urgent action from governments, companies, and developers to ensure that gender equality is a foundational principle in the design, deployment, and governance of AI systems. This proactive approach is deemed crucial to prevent the further entrenchment of inequalities and to build a digital future that is equitable and safe for everyone. The warning comes ahead of major UN-linked AI meetings in Geneva in early July, where these issues are expected to be central to discussions on AI governance.