哪些个人信息不应该直接交给AI? / What Personal Information Should You Avoid Sharing with AI?

Judgment in the Age of AI · Article 12

People often mistake “I am only asking AI to organise this” for “the information remains in my hands”. Once content enters an online service, it may be stored, logged, handled by a provider or governed by another jurisdiction. A safer method is to classify the information before sharing it.

Three privacy circles

Level Examples Response
Outer: publishable Non-sensitive text you would willingly place on a public webpage Still check copyright and information about others
Middle: identifiable Names, contact details, work history, location and account information Remove or replace identifying elements first
Inner: sensitive or irreversible Health, finance, identity documents, passwords, biometrics and children’s data Do not enter into public AI tools

Anonymity is more than deleting a name

A job title, city, unusual event and exact date may still identify someone when combined. Remove names, numbers, addresses, precise times, organisations and unnecessary background. If the task does not require real data, substitute a fictional example.

The ten-second test

If this content appeared before a stranger tomorrow, would I still accept the disclosure? If not, do not give it to a public AI service.

Your duty extends beyond your own data. Do not upload information about colleagues, clients, patients, students or family members without authority.

References


Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.