Data Minimization
Collect and use only the data you actually need—nothing extra, nothing 'just in case.'
Think of it like
Like packing only what fits in a backpack instead of renting a U-Haul for a weekend trip.
Example
A recommendation system needs user behavior, but doesn't need their full browsing history, location, or friendship network—so you don't collect it.
How it actually works
The principle is simple: less data, less risk. Fewer datapoints mean less exposure if there's a breach, less memory load for memorization attacks, and often clearer training signal because you're not drowning in noise. The hard part is resisting the urge to hoarde data 'just in case' it becomes useful later. Privacy regs like GDPR bake this in.
For product teams
Reduce compliance burden and breach risk by being explicit about what data you actually use.
For engineers
Audit your data pipelines to identify and drop unused columns and records.
Related
- Privacy — Related privacy principle.
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