AI Promised to Level the Playing Field. Women May Pay the Highest Price |
AI Promised to Level the Playing Field. Women May Pay the Highest Price
The tools are new. The penalties are not.
EXPERT OPINION BY REBECCA HINDS, REBECCA HINDS, AUTHOR OF YOUR BEST MEETING EVER @REBHINDS
Illustration: Inc; Photo: Getty Images
Every March, Women’s History Month reminds us how hard women have fought to get into rooms that were never built for them. This year, some of that hard-won progress is slipping.
In the years before ChatGPT launched, women’s representation in top-level management inched upward. Since then, it has slid backwards. Women hold just 31 percent of VP-level roles and above today, despite making up 44 percent of the global workforce. Even after controlling for experience, education, occupation, industry, and employer, they’re paid less than men for the same work.
AI was supposed to help fix this. The technology promised to widen access to expertise, chip away at bias in decision-making, and reward what you know over who you know. But technologies never land in fair workplaces. They land in ones bent by status hierarchies, pecking orders, and penalties that don’t fall evenly.
Now, here’s a bitter irony—the same ladder that helped women advance may now be threatened by AI. Credentials gave women a reliable way to move up the org chart: proof of expertise that couldn’t easily be dismissed or hand-waved away. Today, 47 percent of women aged 25 to 34 hold a bachelor’s degree, compared with 37 percent of men. But what made credentials so effective—standardized, codified, and verifiable—is also what makes the underlying work easier for AI to learn. Recent research by Anthropic puts a number on it: Workers in the most AI-exposed jobs are 16 percentage points more likely to be female.
How Anthropic's Claude AI Became a Co-Founder
Built by whom, for whom
AI learns from a crooked past. In a 2025 study of 332,044 real online job postings, researchers found that large language models favored men for higher-paying roles while steering equally qualified women toward lower-paying ones. The models did not invent that bias. They absorbed it from hiring data that had been marinated in lopsided decisions.
When the people designing, training, and auditing these systems are overwhelmingly male, the biases baked into the data are less likely to be questioned, or even noticed. Women make up roughly 22 percent of the AI workforce globally, and only 12 percent of leading AI researchers.
The adoption penalty
Then there is the question of who feels safe picking up these tools. A synthesis of 18 studies found that women adopt generative AI tools at a much lower rate than men. In Denmark, a study of 18,000 workers found women were significantly less likely to use ChatGPT for work, even when doing the same tasks in the same roles. Women were just as optimistic about AI’s productivity potential, but less familiar with the tools and more concerned about whether using them was ethical or allowed. Even when employers banned AI, much of the gap remained. Men were more likely to use the banned tools anyway.