Out-of-Distribution Detection
A technical capability that enables an AI system to identify when input data significantly differs from its training data distribution, indicating conditions where model predictions may be unreliable.
In plain language
The ability of an AI to recognise when it is seeing something very different from what it was trained on and to flag it as uncertain. Like a radiologist saying, "This X-ray shows something unusual that I am not confident diagnosing."
Why this matters
Out-of-distribution detection is a critical safety control that prevents AI systems from making confident predictions in situations they are not equipped to handle. Australian organisations should require OOD detection for all high-stakes AI deployments to meet their duty of care and governance obligations.
Relevance
ImplementationA technical safeguard that prevents high-confidence predictions on novel inputs, reducing operational risk in production deployments.
Related terms
Putting out-of-distribution detection 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.