Data Privacy & Client Confidentiality Basics
What never goes into a prompt, and how to work with sensitive deals safely
Real estate runs on private information
Financial qualification, family circumstances behind a sale (divorce, job loss, estate settlement), off-market listings, negotiation strategy — this is some of the most sensitive information any client shares with you. Before you build AI into your daily workflow, you need a clear, simple standard for what does and doesn’t belong in a prompt.
The core rule
Key principle: If you wouldn’t paste it into a public Slack channel or forward it in an unencrypted email to a stranger, don’t paste it into a chat tool without checking your brokerage’s data policy and the tool’s enterprise/business privacy terms first.
Consumer-grade AI chat tools generally aren’t built to be the system of record for regulated financial data, and your brokerage almost certainly has policies — sometimes contractual ones with your MLS or transaction management software — about where client financial and personal information can live. This lesson isn’t about avoiding AI; it’s about using it on the right layer of information.
Categorize before you paste
| Generally fine to include | Handle with care or anonymize |
|---|---|
| Property features, square footage, price range | Client full legal names tied to financial details |
| Neighborhood and market facts | Social Security numbers, account numbers, loan details |
| Generic buyer criteria (“3BR, under $500K, good schools”) | Reasons for a sale tied to identifiable people (divorce, health, job loss) |
| Publicly listed information (MLS-active listings) | Off-market or pre-market listing details before the seller has agreed to that disclosure path |
| Your own draft communications before sending | Another party’s confidential communications shared with you in confidence |
A simple anonymization habit
Most of the time, you don’t actually need the client’s real name or exact address to get useful drafting help — you need the situation.
Instead of:
"Draft an email to Sarah Martinez explaining why her offer on
742 Wexford Lane was rejected, given her financial situation
after her recent divorce."
Try:
"Draft a warm, professional email to a buyer explaining that their
offer wasn't accepted this round, without going into personal
detail about their circumstances. Keep the door open for future
opportunities. Placeholder name: [Buyer Name]."You get the same quality draft, and you fill in the name and personal touches yourself before sending — which you should be doing anyway as your final review step.
Off-market and pre-market listings
Sellers sometimes want a property quietly tested before it hits the MLS. Treat pre-public listing details the same way you’d treat any client confidence — assume they should not be pasted into any tool until the seller has explicitly agreed to that use, and check with your broker on your brokerage’s specific pocket-listing and data-handling policies.
Compliance warning: Some MLS rules and state laws regulate how and when a property can be marketed before formal listing. Using AI to draft “coming soon” marketing doesn’t change those underlying rules — confirm timing and disclosure requirements with your broker before any pre-MLS marketing goes out, AI-assisted or not.
Building the habit
The goal isn’t paranoia — it’s a five-second pause before you paste. Ask yourself: is this a fact about a property or market (safe), or a fact about a specific person’s life and finances (pause and anonymize or skip)?
Try it: Take your next client-related prompt and rewrite it with placeholders (“[Buyer Name]”, “[Address]”) before sending it to Claude. Notice that the output quality barely changes — you can always fill in real details in your final draft.