Certified Robustness
A provable mathematical guarantee that a model's predictions will not change within a specified perturbation radius around any input, providing formal security assurances.
In plain language
A mathematical proof that an AI will keep giving the same answer even if the input is slightly changed. Rather than just testing examples, certified robustness provides a guarantee that holds within defined boundaries.
Why this matters
Certified robustness provides formal assurance for high-stakes decisions where small input changes should not flip critical outcomes. For organisations deploying AI in safety-critical or security-sensitive contexts, certified robustness may be a required governance safeguard.
Relevance
ImplementationCertified robustness is a technical property relevant to models used in high-stakes decisions where adversarial robustness is a requirement.
Related terms
Putting certified robustness into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.