Can AI make biased or unfair decisions about people?

Yes, it can. And it happens more often than most people realize. AI isn't born biased — it learns bias from us. When you train an AI on data that reflects real-world inequalities, the AI soaks up those patterns like a sponge. It doesn't know they're unfair. It just sees them as normal. A famous example comes from hiring tools. Amazon built an AI recruiting system a few years back. They fed it ten years of resumes. The problem? Most of those resumes came from men, because the tech industry has a long history of being male-dominated. The AI learned that male candidates were preferable. It started downgrading resumes that contained the word "women's" — like "women's chess club captain." Amazon caught it and scrapped the project, but the lesson stuck. This isn't a one-off. Facial recognition systems have been shown to misidentify Black and Asian faces far more often than white faces, according to research from MIT and the National Institute of Standards and Technology. The root cause is almost always the training data. If you train a system mostly on light-skinned faces, it won't understand darker-skinned faces very well. It's not malice. It's a gap in the data. But the real-world harm is serious. People have been wrongfully arrested due to faulty facial recognition matches. Here's what I've found helps when thinking about this: bias isn't a bug you fix once. It's more like a garden you have to keep weeding. You need to check your data, test your models on different groups of people, and have real humans review high-stakes decisions. If a company says their AI is "completely unbiased," that's usually a red flag. No system is. The honest ones admit the risk and work to manage it.