Artificial intelligence is remaking economies, and a report from UN Women presented at the General Assembly puts a number on who is doing the remaking. Women make up 22% of the professionals working in AI, and fewer than 14% of those in policy-making or senior leadership roles in the field. Women are also about 20% less likely than men to use generative AI tools at all.
The three gaps
The report describes a gap that compounds at each stage. Fewer women build the systems. Fewer women are in the rooms where the rules for those systems are decided. And fewer women use the tools once they exist, which means the habits, the shortcuts and the career advantage that come from using them accrue unevenly from the start. Sarah Hendriks, director of policy and programme at UN Women, presented the findings as a warning that the technology could widen inequality rather than narrow it.
Harm, not just exclusion
The second half of the report is about what AI is doing to women who are not in the industry at all. Roughly one in four cases of online violence against human-rights defenders and women journalists now involves misuse of the technology — deepfakes, synthetic imagery, automated harassment. And the International Labour Organization's data, cited in the report, finds that female-dominated occupations are at twice the risk from generative AI as male-dominated ones, because clerical, administrative and support work is what these models automate first.
UN Women's asks are equal participation in AI development, stronger legal protection against misuse of the technology, and international regulation that treats women's safety as part of AI safety rather than a separate subject.
What the numbers look like in Bangladesh
There is no Bangladeshi equivalent of this report, which is itself part of the problem. What is measurable points the same way. Women are a minority of computer-science and EEE intakes at the public universities; the BASIS membership's workforce surveys have long put women at well under a third of the IT industry; and in freelancing — the route into tech that Bangladesh actually scaled — the gap widens rather than closes, because the work depends on late-night client calls and an internationally verifiable payment identity, both harder for many women here to arrange.
The harm side is worse documented and worse handled. Deepfake images made to shame women, especially around elections and family disputes, now arrive at police stations faster than any officer has been trained to assess, and the Cyber Security Act engages only after the material is published. The cheapest interventions are not technological: a women's help desk in each cyber unit that can act on a complaint the same day, and school-level teaching that a convincing image proves nothing. Neither requires a model or a data centre.
Why this belongs in a technology section
Because the 22% is not a social statistic. It is an engineering one. A system built and evaluated by people who share a set of blind spots will carry those blind spots into every product trained on it — and the countries that adopt those products wholesale, Bangladesh among them, inherit the blind spots without ever having been in the room.
