Differential privacy is a way to learn from data about many people while making sure the result reveals very little about any one person.
In plain words
When a model trains on personal data, it can end up remembering details about individuals. Differential privacy adds carefully measured random noise during training so no single person's data changes the result much. The model still learns general patterns, and the privacy protection can be measured.
A quick example
A hospital wants a model trained on patient records. With differential privacy, noise is added during training so the model learns what is common across patients, but nobody could reliably tell from the model whether a particular patient's record was used.
Where you'll meet it
- Training models on sensitive personal data.
- Privacy protection outside machine learning too.
Related
Sources
- Docsdifferential privacy, Google Machine Learning Glossary · read 29 Sept 2026
- PaperDeep Learning with Differential Privacy, Abadi et al. (arXiv) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.