AI is still getting women wrong. The UN has the data.

A new UN Women report came out this week, and I want to sit with it together because it touches every one of us.

On June 22, UN Women published an analysis of 133 AI systems and found that 44 percent showed gender bias. More than a quarter showed both gender and racial bias. When researchers asked large language models to complete a sentence that began with a person's gender, about one in five responses came back sexist. Of the 138 national AI strategies they reviewed, only 18 included substantive gender-responsive measures.

If those numbers make your shoulders drop a little, mine did too.

Jayathma Wickramanayake, who leads UN Women's work on digital technologies, put it plainly. AI models, she said, "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." That's not a glitch. That's the inheritance.

Most of us use AI now — to write emails, plan workouts, draft texts, look up health questions, ease a tough Tuesday. When the tools we use assume "doctor" is a man and "nurse" is a woman, when they soften assertive language only for female-coded names, when their image generations of "CEO" all look the same, those aren't paper cuts. They're tiny daily nudges about who gets to be what. Multiply that across a generation of women growing up with these tools and you can feel the gravity of it.

Women still make up only 30 percent of the global AI workforce. The room where these systems get designed isn't the room you and I live in. That's why the report matters, and why your noticing matters.

One small thing to try today

The next time an AI tool you use says something that feels off — defaults to a male pronoun for a leader, hands you a recipe meant for "him," softens your draft until it stops sounding like you — push back in the chat. Say: that assumed something about gender, try again. Save the screenshot. These corrections become data. Some of them already feed back into the training pipelines. Your "no, try again" is a small vote for what the future model thinks normal looks like.

It's such a small act. But the women who built the field we're now arguing about — Joy Buolamwini, Timnit Gebru, Safiya Noble — they showed us years ago that this is how it changes: women refusing to be misread, on the record.

Quote to sit with

Who codes matters. How we code matters, and why we code matters. — Joy Buolamwini

💌 Sarah