Synthetic data
By Gareth HoyleLast updated How we research
Synthetic data is artificially generated data, produced by models, simulations or rules, that imitates the patterns of real data without being recorded from real events.
Synthetic data is data generated by a model, simulation or set of rules rather than recorded from real people or events. It is used to fill gaps, create rare cases, balance classes and produce training examples where real data is scarce or sensitive.
Can synthetic data replace real data?
Synthetic data cannot fully replace real data. A 2024 paper in Nature by Ilia Shumailov and colleagues found that models trained recursively on data generated by earlier models lose the tails of the original distribution, a process the authors called model collapse. Human-created data remains in demand to anchor synthetic sets.
Is synthetic data free of personal data?
Synthetic data is not automatically free of personal data. If the generator was trained on personal data it can reproduce real records, so the output needs the same re-identification risk testing as any de-identified set. NIST's SP 800-188, published in September 2023, lists generating synthetic data with models among its de-identification techniques, alongside re-identification studies.
If you hold UK data, the same test applies under the ICO's identifiability approach.