Protect the people behind the data

What you’ll be able to do

• Recognize information that needs extra care.
• Reduce a task to the minimum data it actually requires.
• Explain why removing a name may not make a record anonymous.

Start with necessity

Before uploading a file, ask: What does the tool need to accomplish the task? A formatting exercise may require the structure of a document but none of its real personal details. A fictional example may be enough.

Personal information can include names, contact details, identifying numbers, and combinations of details that point to one person. Confidential business information and unpublished work may also need protection, even when they do not identify an individual.

This lesson provides a practical review habit, not a substitute for your organization’s privacy rules or applicable requirements.

A worked example: too much context

Suppose you want help organizing a volunteer schedule. Your source includes names, phone numbers, availability, and notes about why someone cannot attend. The assistant may need availability and assigned roles. It probably does not need phone numbers or personal explanations.

For practice, replace real people with Volunteer A and Volunteer B, omit contact details, and use invented availability. Before doing the real task, check whether your organization has approved the tool and the data use.

Four choices before you share

Minimize: Remove fields that do not affect the task.

Generalize: Replace a precise detail with a broader category when precision is unnecessary.

Substitute: Use synthetic or fictional data to test the workflow.

Withhold: Do not send the material when its sensitivity, permissions, or the tool’s handling are unclear.

These are judgment steps, not a guarantee of anonymity. A rare job title, exact date, and small location may identify someone even after their name is removed.

Try a data-minimization prompt

Help me design a volunteer scheduling template using fictional data only.
Use Volunteer A, Volunteer B, and Volunteer C. Include availability and assigned task.
Do not request real names, phone numbers, personal explanations, or identifying details.
Before producing the template, list the minimum fields needed and explain why each is needed.

Your practice: redesign an input

1. Invent a six-field record for a scheduling task. Include two fields that are clearly unnecessary.
2. Identify the task’s purpose and the minimum useful input.
3. Remove unnecessary fields and replace the remaining values with fictional examples.
4. Ask whether any combination of remaining details could still identify a real person if this were actual data.
5. Write the condition under which you would stop and seek an approved process instead.

Your finished artifact: An original fictional record, a minimized version, and a five-line sharing decision. Do not use actual personal records for the exercise.

Check the service, not just the prompt

A prompt saying “keep this private” does not determine how a provider stores data or who can access an account. Review the product’s current settings, contractual terms, organization policies, retention controls, and available permissions before handling sensitive material. Different account types can have different arrangements.

A common mistake

Believing that permission to read a file automatically means permission to upload it to another service. Access and onward sharing are separate questions.

Reflection

Which field do you routinely include out of habit that your task does not actually need?

Go deeper

NIST’s AI Risk Management Framework places risk management in the context of how systems are designed and used.