Artificial Intelligence Policy Tracker
Search
Menu

AI governance glossary

Algorithmic bias

What is algorithmic bias?

Algorithmic bias is systematic error in an AI system's outputs that produces unjustified differences in treatment or outcomes for people or groups, for example by sex, race, age or disability. It can come from the data, the model's design, or the context in which the system is used.

Why it matters

  • Discriminatory outcomes create liability under existing anti-discrimination, consumer and employment law, regardless of AI-specific rules.
  • The EU AI Act requires providers of high-risk systems to examine training, validation and testing data for possible biases and to take measures to detect, prevent and mitigate them (Article 10).
  • Hiring tools face specific checks, such as New York City's independent bias audit for automated employment decision tools.

In detail

NIST's guidance on managing bias in AI (SP 1270) stresses that bias arises not only in data and algorithms but also in the human and societal context in which systems are built and used, so testing a model's statistics alone is not enough. Many laws address the effect rather than the cause: unlawful discrimination is prohibited whether or not a system was designed to discriminate, and some regimes require bias audits or impact assessments before use.

Source: NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. A plain-language explanation for orientation, not the legal text; follow the source for the binding wording.

Employment AI bias audit kit →

Laws and policies on record that use it

Duties that mention it

  • Bias audit (automated employment decision tools)Under New York City Local Law 144, an impartial evaluation by an independent auditor of an automated employmen...
  • AI impact assessmentAn AI impact assessment is a structured evaluation, before deployment and whenever things change, of how an AI...
  • AI explainabilityAI explainability is the ability of an AI system, or the people operating it, to give reasons for an output th...
  • Red teaming (AI)AI red teaming is structured adversarial testing in which testers try to make an AI system fail, by producing...

All glossary terms · Think a definition is off? Tell us; corrections are logged on the corrections page.

Frequently asked questions

What is algorithmic bias?

Algorithmic bias is systematic error in an AI system's outputs that produces unjustified differences in treatment or outcomes for people or groups, for example by sex, race, age or disability. It can come from the data, the model's design, or the context in which the system is used.

Where does the term algorithmic bias come from?

This explanation follows NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. It is a plain-language paraphrase for orientation; the source has the binding wording.

Which laws and policies use the term algorithmic bias?

Among the records on this site: Colorado AI Act (Colorado) and The UAE Charter for the Development and Use of Artificial Intelligence (UAE).