Amazon scrapped its own AI recruiting tool in 2018 after the system learned to penalize résumés that contained the word “women’s.” The training data reflected years of male dominance in tech, so the algorithm effectively taught itself that female candidates were a worse fit. The company abandoned the project rather than try to fix it. Now, a similar pattern is playing out in a lawsuit against Meta.
That case alleges that Meta’s AI-driven layoff tool ignored legally protected characteristics — specifically, periods of parental or sick leave. The two situations, separated by years and by companies, share a common thread: algorithms that lack context can turn historical bias into automated discrimination.
Think of it like a cook who learns to make soup only from recipes that call for too much salt. Every batch comes out salty, even when a new recipe asks for less. The cook doesn’t know any better.
An AI trained on a narrow or skewed slice of the past doesn’t know any better either. The practical upshot for ordinary workers is that an algorithm can hurt you without anyone at the company explicitly deciding to hurt you.
Bias gets baked in at the data stage — not through a malicious line of code, but through a system that faithfully reproduces the patterns it was fed.
A familiar pattern, different contexts
Amazon’s 2018 tool was trained on résumés submitted over a period, most of which came from men. The algorithm learned to downgrade any résumé that included the word “women’s.” It wasn’t that the company wanted to exclude women. The algorithm just spotted a statistical correlation between that word and unsuccessful applications, and it acted on it.
In the Meta case, the alleged problem involves periods of leave. If an employee took parental leave or sick leave, the AI layoff tool may have treated that as a negative signal — again, not because anyone wrote a rule saying “leave is bad,” but because the historical data showed that people on leave were not around to contribute, and the algorithm picked up on that absence as a predictor of low performance.
Both cases highlight a core tension in using AI for personnel decisions. The technology can surface patterns that humans miss, but it can also surface patterns that humans would never want to use.
What it means for employer liability
Meta’s case may help shape how courts view employer liability for AI-driven personnel decisions. If a judge finds that Meta is liable for its AI layoff tool’s treatment of parental and sick leave, it could set a precedent. Employers would have to verify that their algorithms are not just technically accurate, but socially fair.
That means auditing training data, testing for disparate impact, and building in safeguards that a purely statistical model would never consider. Amazon avoided that legal test by killing its recruiting tool before it ever went into wide use.
Meta is facing it head-on. The outcome could ripple across every company that uses AI to hire, fire, or promote. For now, the lesson for ordinary workers is that the algorithm is only as good as the data it learned from — and that data often carries the biases of the past.
The job of making it fair falls on the people who build it, not the people it evaluates.
Sources
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “**Amazon and Meta AI tools face bias lawsuits over hiring and layoffs**”, “description”: “Amazon scrapped its own AI recruiting tool in 2018 after the system learned to penalize résumés that contained the word “women’s.” The training data reflected…”, “datePublished”: “2026-07-19T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Erika Johnson”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.2bme.xyz/”}, “citation”: [{“@type”: “CreativeWork”, “url”: “https://en.wikipedia.org/wiki/Amazon_%28company%29”}]}
