Remember the federal lawsuit against Workday, or the ACLU complaint over that HireVue screening tool? Those cases argued specific hiring algorithms were discriminating against specific women. This week the story got a lot bigger, and it's not a lawsuit this time — it's peer-reviewed science. A sweeping new study out of Stanford, UC Berkeley Haas, and Oxford's Autonomy Institute, published in the journal Nature, set out to measure something researchers had only studied in narrow slices before: how AI systems represent women at scale, across the entire internet.

The resume finding is the one that should stop every working woman mid-scroll. When the researchers had ChatGPT evaluate resumes, it rated older men more highly than women for the same positions, and that held true whether the researchers used real names or let the AI invent its own job candidates. In other words, this isn't a fluke you can train out with a better prompt. It's baked into how the model was built. If you've ever wondered why a callback didn't come, or why a promotion went sideways after your company quietly rolled out an AI-assisted review process, this is the receipt.

It's not just resumes, either. The team analyzed 1.4 million images and videos from sites like Google, Wikipedia, IMDb, Flickr and YouTube, plus nine large language models trained on billions of words, and found women are consistently pictured and described as younger than men in the exact same high-status jobs. Zoom out further and the pattern holds: a separate UN review of 133 different AI systems found 44 percent showed outright gender bias, and more than a quarter showed both gender and racial bias together. Ask a chatbot to simply finish a sentence about a person, and roughly one in five completions comes back sexist or misogynistic, sometimes describing women as property rather than people.

This is exactly why so many women are hesitant to lean on AI at work in the first place, and new survey data from Lean In backs that up. Women report getting less managerial support for using AI tools, and researchers found their reluctance often isn't stubbornness, it's calculation: they're anticipating harsher judgment for the exact same AI-assisted work a man would get credit for. Falling behind now on a tool this consequential could shape entire careers, which is exactly the trap. Avoid AI because it's biased against you, and you risk falling behind on the skill everyone's being measured against.

My Take Here's the thing — none of this means women should quit using AI or wait for someone else to fix it. It means the fix has to happen upstream, in the training data, the model weights, the audit process, before a resume ever gets scored. And that only happens with more women actually building, auditing, and governing these systems, not just reading studies about how badly they treat us.

If you want a concrete way into that room, Women in CyberSecurity just launched an AI Security Accelerator Program, funded by the state of Maryland, aimed at unemployed, underemployed and mid-career women moving into roles at the intersection of AI and security, exactly the kind of work that decides how these systems get vetted before they touch your resume. Participants get hands-on training from the SANS Institute, plus one-on-one mentorship and personalized career advising. It's Maryland-specific for now, but WiCyS has chapters and programs nationwide, so it's worth checking what's open near you even if this exact cohort isn't.

I won't pretend the regulatory backstop is catching up fast enough to save us in the meantime. As of 2025, 117 countries report some effort to address AI-related digital harm, but progress remains fragmented and the laws lag the technology. Less than 40 percent of countries even have laws addressing basic cyber harassment, let alone the subtler bias baked into a hiring algorithm. That gap is real, and it's exactly why the pressure has to come from both directions at once: women in the rooms where these systems are trained, and women loud enough on the outside that regulators can't keep pretending this is fine.