AIRiskAware
AI Governance Glossary
Governance Concept

What Is Ground Truth?

Ground Truth is the reference data, treated as correct, against which an AI model's predictions are trained and evaluated.

Definition

Ground Truththe reference data, treated as correct, against which an AI model's predictions are trained and evaluated.

Ground truth is the "right answer" a model is measured against — for example, the verified labels in a dataset. Its quality is decisive: if the ground truth is biased, incomplete, or wrong, the model will faithfully learn those flaws, which is why scrutinising how ground truth was created is a key governance question.

Source: Machine-learning practice

Plain-language explanation

Ground truth is the "right answer" a model is measured against — for example, the verified labels in a dataset. Its quality is decisive: if the ground truth is biased, incomplete, or wrong, the model will faithfully learn those flaws, which is why scrutinising how ground truth was created is a key governance question.

Primary source: Machine-learning practice

Related terms

Training Data Governance Data Poisoning Overfitting Bias Testing

See where you stand on AI governance

Take the free 7-question maturity assessment and get a personalised action plan.

Free assessment — 3 minutes →