A new report from the US National Partnership for Women & Families says women may face disproportionate disruption from artificial intelligence in the workforce. The study found that while women make up about 47 percent of the US workforce, they account for 83 percent of workers across 15 occupations identified as having the highest exposure to AI.

Those roles include secretaries, receptionists, and office clerks, among others. According to the report, around 6 million women work in these positions.

Researchers said workers in these jobs may face greater challenges adapting to AI-related workplace changes due to lower access to resources and reduced flexibility in transitioning to other roles. The report also looked at sectors where women are more heavily represented but less likely to face full automation, including nursing, childcare, and home health care, in addition to others.

While these jobs typically require direct human interaction and physical presence, the study said AI could still affect workers in those fields through monitoring and workplace management systems: “These management systems, sometimes described as bossware, can be difficult for workers to understand or challenge, and may worsen job quality even where jobs are not eliminated,” the report said.

The study also raised concerns about women’s representation in AI development and leadership roles. “This underrepresentation means women have less influence over how AI systems are designed, deployed and regulated – even as these systems increasingly shape their work lives.”

Researchers pointed to prior studies indicating that biased training data can influence AI-generated outputs. In one experiment cited by the report, ChatGPT was asked to create resumes using male and female names. According to the findings, resumes generated for women presented them as younger and less experienced, while later evaluations ranked older men’s resumes higher.

The report also referenced research suggesting women may face stronger professional penalties for using AI tools at work.

In one case, participants reviewed identical code samples labeled as either AI-assisted or not. The study found that when a woman was believed to have used AI, the perceived competency penalty was more than twice that applied to a man.