Bias
When an AI's answers unfairly favour or disadvantage certain groups, ideas, or points of view, usually because those patterns were present in the data it learned from. An AI learns from human writing, and human writing carries human prejudices, so the model can absorb and repeat them. This is called bias in AI.
Bias in AI means the system produces results that are skewed in unfair or unrepresentative ways. The root cause is usually the training data. A model learns from vast amounts of text and images made by people, and that material reflects the world as it is, including its imbalances, stereotypes, and blind spots. The model picks these up along with everything else.
The effects can be subtle or stark. An AI asked to picture a “nurse” or a “chief executive” might lean on tired stereotypes. A tool trained mostly on one language or culture may handle others less well. A system used to screen job applications could quietly disadvantage people whose backgrounds were underrepresented in its data. None of this requires anyone to have intended harm.
This matters to you because it shapes the answers you receive. AI can present a narrow or slanted view with the same fluent confidence it brings to everything else, which makes bias easy to miss. Developers work to reduce it through better data and guardrails, but no model is entirely free of it.
The practical response is a thoughtful one rather than a fearful one. Treat AI output as one perspective, not the final word, especially on questions involving people, culture, or fairness. Asking for a range of viewpoints, or checking against other sources, helps you catch a skewed answer before you rely on it.
Related terms
Used in
- AI safety and ethics · Guide