Using AI without leaking private information
Simple habits to protect yourself, your clients, and your team.
AI tools are useful precisely because they read what you give them. That's also the risk. Here are the habits we practice and teach.
Assume your inputs may be used
Free tiers of most consumer AI tools may use your prompts for training. Paid business/enterprise tiers usually don't — check the settings.
Redact before you paste
Replace names, account numbers, and specifics with placeholders. "[CLIENT NAME]" instead of the real name. The AI's answer will still work; you just fill the names back in.
Never paste real credentials
No API keys, no passwords, no full credit-card numbers. Ever.
Ask about client data policy
If you're using AI on behalf of a client or employer, check their policy. Many companies now have an approved tool and rules for what can go in it.
Turn off training when possible
In ChatGPT: Settings -> Data controls -> turn off "Improve the model for everyone." In Claude: it doesn't train on API/paid conversations by default.