Advanced · Safety

Differential Privacy

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. A mathematical framework that bounds how much any single individual's data can affect a model or query result.

Technical Definition

A mathematical framework that bounds how much any single individual's data can affect a model or query result.

How it works

DP adds calibrated noise so that statistics — or trained models — are nearly identical whether a given person's data is included or not. Used by Apple, Google, and the U.S. Census. The privacy budget ε quantifies the trade-off with utility.

Related Concepts

• Federated Learning — Distributed training where models learn from data on many devices without the data ever leaving those devices.
• AI Safety & Alignment — The field ensuring AI systems behave as intended, remain under human control, and avoid unintended harm.
• Privacy — Protecting personal information that flows into or out of an AI system.