Accuracy
The proportion of correct predictions made by an AI model out of the total number of predictions, a fundamental metric for classification tasks.
In plain language
The simplest way to measure AI performance: out of all predictions, what percentage were correct? If the AI made 100 predictions and 90 were right, its accuracy is 90%.
Why this matters
Accuracy is the metric most often quoted and most often misread. A model that is 99% accurate can still be useless or harmful if the 1% it gets wrong falls disproportionately on one group of people or on the high-stakes cases that matter most. For governance, the lesson is that headline accuracy must never be reported in isolation: pair it with measures such as precision, recall and performance broken down by customer segment, so decision-makers understand real-world reliability instead of a single flattering number.
Relevance
GovernanceAccuracy is a governance concern because it is the headline number leaders most often rely on and most often misread. Oversight means insisting it is reported alongside other measures so the board sees true reliability.
Related terms
Putting accuracy 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.