Avoid Sensitive Data: What Not to Paste Into AI Tools at Work

AI tools are powerful precisely because they accept any text you give them. That same openness creates a data exposure risk that most sales reps don't think about until it becomes a problem. Understanding what not to paste into a public AI tool is a basic professional hygiene requirement — not a legal technicality.

What counts as sensitive in a sales context:

• Customer contract terms, deal values, and pricing specifics tied to named accounts
• Non-public company financial information (your own or the prospect's)
• Personal contact information beyond what's publicly available (direct cell numbers, personal email addresses)
• Content from NDAs or contracts marked confidential
• Internal forecast data and pipeline numbers tied to specific accounts

What's generally fine to use:

• Publicly available company information (website content, press releases, public earnings call transcripts)
• Generalized discovery notes that don't identify the account by name
• Your own sales materials, value props, and playbooks
• Generic persona and ICP descriptions

The safer path for sensitive workflows: If you need AI to work with sensitive account data, use a platform with enterprise privacy guarantees — Claude for Enterprise, ChatGPT Enterprise, or Microsoft Copilot (which is tenant-scoped and permission-respecting). These tools don't use your inputs to train their models and provide contractual data handling commitments.

For ChatGPT specifically: turn off "Improve the model for everyone" in Settings → Data Controls. This prevents your inputs from being used for training. It takes 30 seconds and should be the first thing any rep does before using ChatGPT for work.