Data provenance is the verifiable record of where a dataset came from and every transformation it passed through on the way to its current form.
It answers questions a log alone cannot: which source produced this value, what cleaning or aggregation changed it, and who or what touched it last. Strong provenance turns “we think this data is right” into something a team can actually show.
For AI it becomes the thread that connects a model output back to the precise data state that produced it, which is what makes a result auditable instead of a guess when someone asks how the model reached it.