Risk & Safety
How to Use AI Without Putting Customer Data at Risk
A short, practical guide every Delmarva business owner should read before turning AI loose on real customer data.
By Delmarva AI · June 17, 2026
AI tools are not all the same. What you can safely paste into ChatGPT is very different from what you can safely paste into a HIPAA-eligible scribe with a signed BAA.
Rule of thumb
If you would not email it to a stranger, do not paste it into a public AI tool. That includes:
- Social security numbers
- Full credit card numbers
- Patient health information
- Privileged legal communications
- Bank account and routing details
- Children's personal information
What is generally safe
- Marketing copy and ideas
- Public company information
- Your own SOPs and internal notes (if you're fine with them training a model — check the setting)
- Reviewing your own writing
Turn off training
Most paid tiers let you turn off "model training on your data." Do it. It's usually one toggle.
Pick the right tier for the data
- General work: ChatGPT Plus or Claude Pro with training disabled.
- Health data: Only HIPAA-eligible tools with a signed BAA.
- Legal data: Tools with explicit confidentiality terms — or self-hosted.
- Financial data: Same rule — explicit data protection terms or skip it.
Write a one-page AI policy
Even for a team of three. Cover: which tools are approved, what data is off-limits, who to ask if you're unsure. We can help — that is what the AI Readiness Audit is for.
Get the next one in your inbox
The Delmarva AI Brief — one short email, practical AI you can use this week.