Reese Witherspoon, the actor and producer, recently faced backlash after saying, “It’s so, so important that women are involved in AI . . . because the change is here.” Her comments struck a nerve. One response, which received more than 4,300 likes, argued: “The ‘it’s here to stay’ mentality encourages us to dampen our critical thinking skills and accept circumstances that can hurt us.”
Witherspoon’s advice and the resulting pushback reflect the broader divide in how people view artificial intelligence. Some see it as the exciting future of technology, while others are sounding an alarm about its risks. New research from Pew suggests women are more likely than men to fall into the latter camp. They are less likely to believe AI will have a positive impact on their own lives over the next 20 years (17% versus 29%) and more likely to believe that it will have a negative impact on society (43% versus 36%).
Before dismissing women’s concerns as resistance to innovation, leaders should ask what they reveal. Research suggests they are not rooted in a fear of change but in understanding the unequal risks AI presents. Listening to those concerns is essential for leaders who want to deploy AI responsibly.
AI reinforces existing inequities
Rather than leveling the playing field, AI can replicate workplace biases. In an analysis of 133 AI models, nearly half exhibited gender bias, and more than a quarter demonstrated both gender and racial bias. Large language models (LLMs) often reinforce gender stereotypes by associating women with domestic responsibilities, caregiving, and subordinate or sexualized roles while connecting men with leadership, business, and professional achievement. And attempts to make AI more inclusive can also introduce new distortions. One study found LLMs may overcorrect for bias by attributing stereotypically male statements to female authors more often than would occur in reality.
These biases stem from how AI works. LLMs do not generate new knowledge. Instead, they predict each word fragment based on enormous collections of preexisting text, prioritizing linguistic fluency. Because datasets often reflect historical and societal biases, AI can reproduce and amplify them. While an AI sentence may be well written and sound confident, it may be factually incorrect and perpetuate stereotypes.
Bias is embedded not only in AI systems themselves but also in who gets the opportunity to learn and use them. In a survey of 12,000 employees around the world from the recruiting firm Randstad, 71% of men reported having AI expertise. But among women, the number dropped to 29%. While some of the gap may be related to men’s general overconfidence, more men (41%) than women (35%) reported being provided access to tools; men also had more training opportunities (38% versus 33%). And that all leads to a gender gap of feeling prepared to utilize AI at work—35% of men versus 30% of women.
Ironically, when women do use AI, research shows they may receive competence penalties. Zehra Chatoo, an independent researcher, created an AI-supported résumé and asked 1,000 adults in the U.K. to evaluate the candidate. The evaluators received identical résumés and were told that the candidate had used AI. The only difference was the candidate’s name. Half of the evaluators saw Emily Clarke, while half saw James Clark. The evaluators who attributed the AI-assisted résumé to a woman were twice as likely to question the candidate’s competency. “She can’t even write a CV herself—not sure she has the skill to carry out the job,” said one of the evaluators. James’s résumé had a different response, with his use of AI justified: “He just needed a bit of help putting it together.”
In another study, researchers asked 1,026 software engineers at a company to evaluate an identical piece of computer code. The engineers were randomly told that the code had been written by either a man or a woman and with or without AI assistance. While male AI coders received 6% lower competence ratings than non-AI coders, female AI coders received 13% lower competence ratings. Leaders may assume that use of AI is neutral; women’s experiences suggest otherwise.
AI shifts disproportionate risks onto women
Beyond bias, the risks of AI are not evenly distributed. Privacy is one example. Because generative AI systems can retain user inputs, it is often unclear what information is stored, how it may be used, or how to prevent one’s data from training the AI model. That uncertainty can make sharing sensitive information feel risky. These concerns are reflected in how women use AI. Survey research has found that women are significantly less interested than men in relying on generative AI for personal matters like finances, relationships, or health, suggesting they are more sensitive to entrusting AI to handle deeply personal information.
Al also creates threats to women’s safety. Deepfakes, which are AI-generated media that mimic a person’s appearance or voice, take online misogyny to another level. Recent studies estimate that 98% of all deepfake content online is nonconsensual and pornographic, and that 99% of those depicted are women.
Women in public-facing roles are especially vulnerable to online abuse designed to intimidate, silence, and drive them out of public life. In a recent global study, 45% of women journalists reported self-censoring because of online violence. This threat is likely to grow as deepfake tools become user friendly, generate images more quickly, and are increasingly low cost or even free.
AI also carries the risk of alienating customers. When AI voice assistants or chatbots don’t have the answer, who steps in? Human workers. While AI may answer routine questions, specialized or customer-specific situations require human judgement. And many people still prefer talking to a person rather than a chatbot. Customer-facing roles, which are 70% women, can become even more emotionally demanding. Someone must soften AI-generated messages, resolve problems, and repair relationships when the technology falls short. This emotional labor—the often invisible and undervalued work of producing positive feelings—is essential. Yet the burden rarely falls to AI designers or executives, roles overwhelmingly held by men. Instead, the work of restoring customer trust and mitigating AI’s shortcomings disproportionately is carried by women.
AI redistributes economic and environmental costs
Though the number of AI jobs has doubled since 2023, women were just 26% of U.S. AI hires in 2025 and hold just 13% of C-suite AI roles at AI companies across 27 countries. While women are underrepresented in jobs designing AI, they are overrepresented in jobs most exposed to automation. Female-dominated occupations, such as administrative assistants and customer service representatives, are nearly twice as likely to be affected by generative AI as male-dominated occupations such as construction and manufacturing. In many of these female-dominated jobs, AI is being promoted as a tool to automate work rather than augment workers’ capabilities. AI is also shrinking the entry-level positions that provide the first step into professional careers. As women now make up the majority of college graduates in the United States, the loss of these entry points could disproportionately limit their opportunities to gain experience, build networks, and progress into leadership.
AI also carries environmental costs that are often overlooked. A single AI prompt consumes roughly 10 times the energy of a web search. Data centers require enormous amounts of electricity and water, often relying on fossil fuels for power and hundreds of thousands of gallons of water each day for cooling. Public concern reflects these tradeoffs: Women are significantly more likely than men to oppose new data center construction (55% versus 43%) and worry about AI’s climate impacts.
AI’s rapid growth is also increasing demand for electricity and computing hardware, contributing to rising costs for both. Manufacturers are prioritizing memory chip production for data centers, which buy at scale and secure contracts to monopolize future purchases. The memory chip shortage is causing a consumer “AI Tax” for electronics, laptops, smartphones, and even cars. Because women continue to earn less on average than men and are overrepresented in lower-paying occupations, these price increases consume a larger share of their income and leave fewer resources for savings and retirement.
Women aren’t resisting AI—they’re recognizing risks that many organizations have yet to address. But AI is not going away. The challenge for leaders is not whether employees should use it, but how to help them use it responsibly without amplifying the unequal risks women already face. These seven recommendations can help leaders strike that balance.
- Offer holistic AI training. AI workshops often focus on how to use AI and its productivity gains. Ensure that training explains how generative AI works, where it can produce biased or inaccurate outputs, how it may disadvantage women and other underrepresented groups, the competence penalty women can face, and the ethical, privacy, and environmental implications of its use.
- Report AI bias. Encourage employees to report biased, discriminatory, or harmful AI outputs. Escalate these issues to AI vendors, particularly when there are contractual relationships that give leverage to demand improvements.
- Establish transparency guidelines. Develop policies for AI use that address transparency. In what circumstances should an employee disclose their use of AI? Perhaps it is overkill to disclose when AI is used to rewrite an email based on a person’s ideas. But if employees use AI to analyze data or produce a report, then transparency could be warranted. Level the playing field with transparency requirements that apply to everyone.
- Give employees discretion in how they use AI. Due to its nonthinking nature, AI doesn’t make work better on its own. It requires knowledgeable oversight and human judgement to identify errors, verify facts, and detect bias. And when employees become overly reliant on AI, it can erode their critical thinking and problem-solving skills. Set clear performance goals for your staff and let them decide how to best accomplish them. If they choose to use AI, ensure they have checked the work for accuracy, bias, and appropriateness.
- Don’t replace employees with AI. It is tempting to replace human labor with AI. But companies that made cutbacks are reversing course. One bank replaced 40 customer service positions with an AI voice assistant, only to experience a surge in call volume. The bank is now rehiring staff. Even if AI can replace portions of the entry-level employee work, keep the positions. Without those employees your organization will lose a qualified pipeline to promote up the ranks.
- Reward emotional labor. Make the relational work of addressing AI inadequacies visible by recognizing and rewarding it in performance evaluations, promotions, and compensation decisions.
- Flip it to test it. If you question whether a woman’s résumé or other work is AI-generated, consider how you would react if a man submitted identical work. Are you viewing AI as evidence that she lacks competence, while viewing it as a productivity tool when used by a man? Use this check to guard against imposing a competence penalty on women for using AI.
AI is reshaping the workplace, but organizations still have choices about how they implement it. Women’s concerns should not be dismissed; they offer valuable insight into the risks AI can create without sufficient safeguards. Leaders who listen will be better positioned to deploy AI in ways that reduce bias, protect privacy and safety, and preserve critical thinking. Rather than maximizing AI adoption, leaders should focus on maximizing human potential.
