Privacy
The right to keep your data and identity away from prying eyes—including the AI building on it.
Think of it like
Like the difference between a journal you write and one you share with the world.
Example
A model trained on hospital patient data should never memorize and leak someone's medical history in its outputs.
How it actually works
Privacy in AI has two angles: protecting user data during training (not memorizing specific records) and protecting user inputs from being logged or sold. Models can leak training data, especially if trained on small or sensitive datasets. Privacy-preserving techniques like differential privacy add noise so no single person's data can be fully recovered, but you trade accuracy for safety.
For product teams
Build trust by guaranteeing users their data won't be memorized or resold.
For engineers
Use differential privacy, federated learning, or secure aggregation to prevent data leakage during training and inference.
Related
- Differential Privacy — Complementary privacy technique.
- Federated Learning — Decentralized training approach.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome