Artificial intelligence is rapidly becoming part of everyday life; however, as AI becomes more embedded in society, concerns are growing about its ability to reproduce and amplify existing gender stereotypes.
Recent research suggests AI systems can routinely reproduce gender bias because they are trained on historical, human-generated content that reflects decades of unequal representation.
AI reproduces existing gender stereotypes
UN Women found that 44% of 133 AI systems analysed demonstrated gender bias, while 26% showed both gender and racial bias.
Large language models have also been found to associate women with concepts such as "home", "family" and "children", while associating men with "business", "executive", "salary" and "career".
In one study, around 20% of responses from large language models to prompts beginning with a person's gender demonstrated sexist or misogynistic attitudes.
These findings suggest that AI bias is not simply a technical glitch. It can reflect and reproduce patterns of inequality already present in the data used to train these systems.
When bias becomes automated
The concern becomes greater as AI is increasingly used in areas such as recruitment, advertising, search and decision-making.
Bias that might once have existed in individual human decisions can become embedded in automated systems and reproduced at scale.
UN experts have warned that the rapid expansion of AI and digital technologies, without meaningful regulation, could deepen existing gender inequalities and create new risks for the human rights of women and girls.
The UN Working Group on discrimination against women and girls has highlighted the need for gender-responsive governance, warning that AI systems designed without recognising gendered risks could reinforce harmful stereotypes and create new forms of exclusion across political, civic and economic life.
AI is also intensifying online violence
AI is creating new tools for harassment, manipulation and image-based abuse, with women and girls disproportionately affected.
Women human rights defenders, activists and journalists are already experiencing AI-assisted online violence. UN Women reports that almost one in four surveyed women in these groups had experienced AI-assisted online violence.
Deepfakes are one of the most visible examples. AI-generated and manipulated images and videos can be used to harass, intimidate and sexually exploit women without their consent.
As generative AI becomes more accessible, the scale and speed at which this type of abuse can be created and distributed is also increasing.
Who is building AI?
Another concern is who gets to shape the technology in the first place.
Women remain significantly underrepresented in STEM and AI, making up only 30% of the global AI workforce.
This lack of representation matters. The people developing AI systems are helping determine how these technologies are designed, tested and governed, yet they do not fully reflect the billions of people who will use and be affected by them.
Greater diversity in the AI workforce is therefore not only a question of representation. It is also an important part of identifying risks, challenging assumptions and building systems that work for everyone.
AI needs to serve equality, not discrimination
UN experts are calling for stronger, gender-responsive governance and greater accountability to ensure AI does not reinforce discrimination.
The challenge is not to stop AI from evolving, but to ensure it evolves in a way that supports equality rather than automating inequality.
As AI becomes increasingly influential in workplaces, economies and everyday life, addressing gender bias must be part of the conversation from the design stage, not an afterthought once the technology is already embedded in society.