What Is Robustness?
Robustness is an AI system's ability to maintain its performance under varied, unexpected, noisy, or adversarial conditions.
Robustness — an AI system's ability to maintain its performance under varied, unexpected, noisy, or adversarial conditions.
A robust system does not fall apart when the real world differs from its training data or when someone deliberately tries to confuse it. The EU AI Act pairs robustness with accuracy and cybersecurity as a requirement for high-risk systems (Article 15), and the NIST AI RMF treats it as a key trustworthiness characteristic. Brittleness — strong performance in testing that collapses in deployment — is a common AI failure mode.
Source: EU AI Act, Article 15; NIST AI RMF
Plain-language explanation
A robust system does not fall apart when the real world differs from its training data or when someone deliberately tries to confuse it. The EU AI Act pairs robustness with accuracy and cybersecurity as a requirement for high-risk systems (Article 15), and the NIST AI RMF treats it as a key trustworthiness characteristic. Brittleness — strong performance in testing that collapses in deployment — is a common AI failure mode.
Related terms
See where you stand on AI governance
Take the free 7-question maturity assessment and get a personalised action plan.
Free assessment — 3 minutes →