Synthetic Data
Synthetic data is artificially generated data produced by algorithms and generative models rather than collected from real-world events. It is designed to mirror the statistical properties and correlations of a genuine dataset, so a model trained on it learns the same patterns without any real customer record being exposed. It is also used where real examples do not exist in sufficient volume.
In practice
Synthetic data lets a team train and test while keeping personal data inside the boundary the regulator expects, and it fills gaps where real examples are scarce, such as rare fraud cases. It does not remove the obligation: the generator is itself trained on real data, so ask what it saw, who signed off the re-identification risk, and how closely the synthetic set tracks the real distribution.