Checklist

What not to paste into AI tools at work.

A practical safety filter for office workers before using AI.

Why this matters

Most AI anxiety starts with not knowing what is safe.

This page gives the user a simple pre-flight check before they paste anything into an AI tool. It is deliberately conservative for people using public or unclear workplace tools.

Use today

Rewrite the task without the real details.

If the AI can help with the pattern of the work, it does not need the real customer, employee, project, contract, or financial details.

Use this before your first prompt.

Remove first
  • Customer, client, patient, or employee names.
  • Financial figures, contracts, internal strategy, or legal details.
  • Passwords, API keys, personal identifiers, or private messages.
  • Anything your workplace has not approved for AI use.
Safer substitutes
  • Use placeholders such as [customer], [project], or [date].
  • Describe the type of task instead of pasting the real document.
  • Use public or generic examples while learning.
  • Ask AI to help create a safe prompt before sharing context.

If your workplace rules are unclear, use the conservative default.

1

Use approved tools only.

If you do not know whether a personal AI account is acceptable for work content, assume it is not.

2

Remove real details first.

Practise with placeholders, invented examples, public information, or a description of the task pattern.

3

Check before sharing.

Verify names, numbers, dates, policy claims, links, and recommendations before you send or rely on the output.

4

Ask where review happens.

For anything affecting people, money, customers, compliance, or reputation, identify the human reviewer before using AI output.

Copy-paste prompt: make this task safe for AI practice.

# ROLE You are my workplace AI safety filter. # OBJECTIVE Help me turn this work task into a safer AI practice version without using sensitive details. # TASK [describe the task in general terms] # WHAT FEELS UNCLEAR [approved tool / data boundary / review standard / manager expectation / not sure] # PROCESS 1. Identify information I should remove before using AI. 2. Replace risky details with safe placeholders. 3. Suggest a low-risk version of the task I can practise. 4. Tell me what needs human review before I use the result. 5. Give me one safer first prompt. # OUTPUT Return: - Remove first - Safer replacement - Practice version - Human review needed - Safer first prompt
Choose a low-risk workflow Handle unclear AI policy Check the answer Try a safe email prompt