AI prompts for keeping customer data safe.
For anyone on the team who uses AI with business data. These prompts get the answer without handing AI your customers’ data. 6 copy-paste prompts for Claude, ChatGPT, or any AI assistant.
How to use these
Fill in the brackets.
Each prompt has [bracketed] spots. Replace them with your real numbers and context. The more specific you are, the better the answer.
Use a project.
In Claude or ChatGPT, put your policies, product list, and chart of accounts in a project once, so every prompt starts with your context.
Leave customer data out.
AI tools and vendors leak regularly (see the latest incidents). Remove customer names, addresses, emails, and messages before pasting. Card numbers and passwords never go into a chat. Amazon buyer data, including personalization text, has extra rules.
Check the math.
AI is good at structure and explanation and can still get arithmetic wrong. Spot-check any number you will act on.
Keep customer data out
Plan an analysis without sharing customer data.
I want to answer this question with AI: [your question, for example which products repeat buyers order most]. The data comes from [Shopify / Amazon / QuickBooks / our CRM]. These are the column names only, no real rows: [paste column names] Do not ask me for real customer records. Sort every column into Keep (needed and not sensitive), Replace (needed to tell rows apart but sensitive, replace with a code like Customer 001), or Delete. Treat names, emails, phone numbers, street addresses, Amazon buyer details, payment details, and anything that looks like a password or key as sensitive. Then give me step-by-step instructions to clean the file in [Excel / Google Sheets] on my own computer before I share anything.
Write a local script that masks customer columns.
Write a short [Python / PowerShell] script I will run on my own computer. It reads [file name].csv, deletes these columns: [list], replaces these columns with a stable code per unique value (for example Customer 001): [list], and writes [file name]-safe.csv. It must not upload, email, or print any original values. Add comments so a non-developer can follow it, and tell me how to check the output before I share it.
Rewrite an email so it is safe to paste.
I will describe a customer email instead of pasting it. Customer situation: [describe the issue in your own words, with no names, emails, phone numbers, addresses, or order details that identify the person]. Our policy: [paste the relevant policy]. Draft a reply I can personalize myself before sending. Use [Customer] where the name goes.
Check a tool before you use it
Vet an AI tool from its public documents.
Act as a careful IT and security reviewer for a small business. Review this AI tool from its public documents only. Tool: [name and link] What we want to use it for: [use] Data it would touch: [email / files / store / Amazon Seller Central / accounting / CRM] Privacy policy, terms, and security page: [paste text or links] Answer yes, no, or unknown for: trains on our inputs by default, can training be turned off for the company, how long inputs are kept, independent audits (SOC 2 Type II, ISO 27001), where data is stored, data processing agreement available. Then give a verdict (approve, approve with limits, do not use), the data we may and may never put in it, and the settings we must change. Treat missing information as a finding.
Decode what a browser extension or connector can access.
Here are the permissions an AI [browser extension / connector / OAuth app / MCP server] asks for: [paste the permission list or scopes] Explain in plain language what each permission lets it see or do, including on sites like Amazon Seller Central, our bank, and email. Rate the overall risk, say which permissions we could refuse or narrow, and what a safer setup would look like.
Write a one-page AI use policy for the team.
Write a one-page AI use policy for a [size]-person [industry] business. Cover: approved tools, data that is never allowed in AI (customer personal data, Amazon buyer data, payment details, passwords and API keys, anything marked confidential), how to prepare data safely, what to do if someone pastes sensitive data by mistake, who approves new tools, and how often the policy is reviewed. Plain language, no legal jargon, short enough to read in three minutes.
Why these exist
AI tools and vendors leak more often than most teams expect. See the latest incidents. Every prompt here gets the work done without real customer records, Amazon buyer data, payment details, or passwords in the chat.
We hold ourselves to the same rule. See how we build automation and our security audits.
Audited every year.
Deloitte and Amazon SP-API audits in 2026. PwC and Amazon SP-API in 2025. Shadow AI was involved in 43% of the breaches IBM studied for its 2026 report; we keep your data out of it. And if something we run goes wrong, we own the fix.