A company is strengthening its governance for generative AI. What is the name of the recently emphasized concept of recording the usage history of generative AI (when, who, and how a model was invoked) so that it can be audited later?
Select the concept of recording the usage history of model invocations.
Data lineage
Data lineage is a concept that tracks the origin and transformation history of data. Lineage means 'ancestry or provenance,' referring to the state in which one can trace, from the very beginning, where data was obtained and what processing and transformations it went through to reach its current form.
The object of tracking is the provenance of data, not a record of how a model was invoked (who used it, when, and how), so it is incorrect.
Model card
A model card documents a model's intended use, performance, and risks.
It is documentation describing the model, not a record of usage history for each invocation, so it is incorrect.
Data classification
Data classification is the practice of dividing data by sensitivity and labeling it.
It is a rule division for handling, not a record of usage history, so it is incorrect.
AI audit log
This is correct. An AI audit log is a concept that records usage history (when, who, and how a model was invoked) so that it can be audited and traced later. It supports accountability and governance.
Remember the correct answer, the 'AI audit log (record of model invocations).'
- A concept that records the usage history of generative AI (when, who, and how a model was invoked) so it can be audited and traced later.
- It supports accountability and compliance and has been emphasized recently.