Safety & Ethics
Glossary
Bias
Systematic unfairness in a model's outputs, usually inherited from its training data or design choices.
Last updated
Bias appears when a model treats groups or ideas unfairly because its training data or objectives encode historical imbalances. It can affect hiring, lending, healthcare and moderation decisions. Mitigation involves diverse and audited data, testing across groups and keeping humans accountable for high-stakes outcomes.
Related terms
Keep exploring
Browse the full AI glossary or compare AI tools that use this technology.