accuracy
This glossary entry explains accuracy for AI governance and model risk programs. The sections below summarize what the term means in plain language, why chief AI officers and cross-functional committees track it, where teams often get confused, and—when you are signed in—how it shows up across major industries and in expectations tied to the EU AI Act and NIST AI RMF. Use related links at the end of the page to explore neighboring concepts without losing context.
What It Means
Accuracy measures how often your AI system gets the right answer compared to what actually happened or what the correct answer should be. It's calculated as a percentage - if your model makes 100 predictions and 85 are correct, your accuracy is 85%. Think of it as your AI's report card showing how reliable its decisions are in real-world situations.
Why Chief AI Officers Care
Accuracy directly impacts business outcomes and customer trust - low accuracy means bad decisions, unhappy customers, and potential regulatory issues. In regulated industries like healthcare or finance, accuracy requirements may be mandated by law, and falling short can result in fines or license revocation. Poor accuracy also wastes resources as teams spend time fixing mistakes the AI should have caught.
Real-World Example
A bank's fraud detection system reviews 10,000 transactions daily and correctly identifies 9,200 as legitimate or fraudulent - giving it 92% accuracy. The remaining 8% includes both missed fraud (letting bad transactions through) and false alarms (blocking good customers), both of which cost the bank money and damage customer relationships.
Common Confusion
People often think accuracy is the only metric that matters, but a model can have high overall accuracy while still performing poorly on the cases you care most about. For instance, if only 1% of transactions are fraudulent, a lazy model that labels everything as 'not fraud' would still achieve 99% accuracy while catching zero actual fraud.
Industry-Specific Applications
See how this term applies to healthcare, finance, manufacturing, government, tech, and insurance.
Healthcare: In healthcare AI, accuracy is critical for patient safety and must meet FDA standards for medical devices, with diagnost...
Finance: In finance, accuracy is critical for AI models used in credit scoring, fraud detection, and regulatory reporting, where ...
Premium content locked
Includes:
- 6 industry-specific applications
- Relevant regulations by sector
- Real compliance scenarios
- Implementation guidance
Technical Definitions
NISTNational Institute of Standards and Technology
"Closeness of computations or estimates to the exact or true values that the statistics were intended to measure."Source: OECD
"A qualitative assessment of correctness or freedom from error."Source: FDA_Glossary
"The measure of an instrument's capability to approach a true or absolute value. It is a function of precision and bias."Source: FDA_Glossary
"The accuracy of a machine learning system is measured as the percentage of correct predictions or classifications made by the model over a specific data set. It is typically estimated using a test or "hold out" sample, other than the one(s) used to construct the model. Its complement, the error rate, is the proportion of incorrect predictions on the same data."Source: Raynor
"measure of closeness of results of observations, computations, or estimates to the true values or the values accepted as being true"Source: ISO/IEC_TS_5723:2022(en)
Explore more glossary terms
Discuss This Term with Your AI Assistant
Ask how "accuracy" applies to your specific use case and regulatory context.
Start Free Trial