Safe Use Tutorial · 6 Min Read

What Not to Share with AI: The Practical Privacy Guide

When you type into an AI chat window, where does that text go? Learn the essential boundaries between safe inputs and risky disclosures at work and home.

Where Does Your Prompt Data Go?

By default, many consumer-tier AI services retain user prompts and generated responses to train future models, review safety violations, or improve system performance. Unless your organization has an enterprise agreement with zero-data-retention guarantees, you must assume your input could be stored and reviewed.

The “Never Paste” Red List

Never enter these four categories of information into a public consumer AI tool:

• Personally Identifiable Information (PII): Full names, Social Security or national ID numbers, home addresses, personal phone numbers, and passport scans.
• Credentials & Secrets: Passwords, API keys, private cryptographic keys, database connection strings, and security tokens.
• Confidential Business Data: Unannounced financial earnings, pending merger documents, trade secrets, proprietary algorithms, and legal settlement drafts.
• Protected Health Information: Medical diagnoses, patient records, and health insurance claim numbers.

How to Sanitize Prompts Using Placeholders

You can still get 100% of the value from AI writing and analysis tools without risking confidential data. Simply replace real entities with bracketed placeholders:

Check Your Workplace AI Policy

Most organizations now maintain an approved AI usage policy. Always verify whether your company provides an enterprise-licensed workspace (such as ChatGPT Enterprise, Claude Team, or Microsoft Copilot) where model training on prompt data is disabled by contract.