Before You Put Your Health Into an AI Chatbox

AI can help you understand. Do not give it more of your medical life than it needs.

Patients are already using AI for health questions.

I hear it in clinic. Someone gets lab results back and asks a chatbot what "high" means. Someone reads an MRI report and wants the medical language translated into English. Someone takes a picture of a rash, a wound, or a swollen foot and asks whether it looks serious. Someone copies a portal message because they are not sure what the office is asking them to do.

I understand the impulse.

Medicine is hard to read from the outside. The portal gives you numbers, flags, fragments, and messages that often assume you already know the context. Search engines can make everything sound either harmless or catastrophic. Appointments are short. Follow-up can be slow. So a tool that answers immediately feels useful.

Sometimes it is useful.

AI can explain a term. It can help you prepare questions. It can turn a confusing report into something you can discuss with your doctor. It can help you organize your thoughts before a visit.

The problem is that a general AI tool is not the same thing as your doctor's office.

That sounds obvious until you are tired, worried, and staring at a red lab value at 10:30 at night.

A safer use is something like: "What does creatinine mean?" or "What questions should I ask my doctor about a high A1c?" Those are general questions. You are trying to understand a concept.

A different use is pasting your full lab report with your name, date of birth, medical record number, doctor, clinic, and insurance information still attached.

That second version is no longer just a question. It is a piece of your medical record.

The same thing happens with visit notes, medication lists, portal threads, discharge instructions, imaging reports, and photos. Patients may think they are just asking for help. But the tool may be receiving names, dates, locations, rare diagnoses, faces, body images, family details, medication histories, or identifiers that were never needed for the question.

This is where people need a simple rule:

Do not give an AI tool more of your medical life than it needs.

If you want help understanding a lab value, ask about the lab value. You usually do not need to paste the entire report.

If you want to understand a diagnosis, ask generally first. "What is peripheral neuropathy?" is safer than uploading a full clinic note.

If you want help writing a message to your doctor, draft it without copying the whole portal thread. You can say, "Help me ask whether this medication side effect needs follow-up," without including names, dates, and every detail from the chart.

Photos deserve extra caution. A picture can reveal more than you think. It may show a face, a distinctive tattoo, a room, a timestamp, a wound location, or a body part you would not want stored outside your care team. Even when the photo is cropped, it may still be sensitive. A rash photo, wound photo, surgical photo, or X-ray screenshot is not just an image. It is health information.

Medication lists are also tricky. A single medication may not identify you. A long list, combined with age, sex, location, diagnosis, or a rare condition, can become more identifiable. Insurance IDs, addresses, phone numbers, Social Security numbers, medical record numbers, and full dates of birth should not go into a general AI tool.

None of this means patients should avoid AI completely.

That advice will not work. People are already using it because the need is real. The better advice is to use it like a translator or question-prep tool, not like a place to store your chart.

Good uses:

• "Explain this medical term in plain English."
• "What questions should I ask my doctor about this diagnosis?"
• "What does this lab test usually measure?"
• "Help me write a short message asking whether I need follow-up."
• "What symptoms would generally make this more urgent?"

Riskier uses:

• Uploading a full report without removing identifiers.
• Pasting a portal thread with names and dates.
• Uploading photos of wounds, rashes, faces, or body parts.
• Sharing insurance information, addresses, or full medication histories.
• Asking AI to decide whether you should skip care.

That last one matters.

AI can help you prepare for a conversation with a clinician. It should not be the reason you delay care when something is changing. It cannot examine you. It does not know your full history. It cannot feel the skin temperature around a wound, notice the way someone is walking, review the outside records that never made it into the portal, or take responsibility if it reassures you incorrectly.

Use it to become a better question-asker. Do not use it as a substitute for being seen.

Healthcare privacy rules were built around protected health information: who receives it, who stores it, who can use it, and what safeguards are required. HHS guidance on cloud computing makes clear that healthcare organizations can use cloud services, but electronic protected health information still needs appropriate safeguards and business associate responsibilities. That is a very different setup from a general consumer tool used casually at home.

There are healthcare-specific AI systems with stricter controls. Some may have agreements about privacy, retention, audit logs, and permitted use. But patients should not assume every tool that can answer a health question was built for healthcare.

That is the trap.

The tool may sound calm. It may sound medically fluent. It may even give a good explanation. But sounding helpful does not tell you what happens to the information you typed.

Before putting health information into any AI tool, ask yourself:

Would I be comfortable with this being stored?

Would I be comfortable with someone outside my care team seeing it?

Can I ask the same question without my name, date of birth, location, photo, or full report?

Am I using this to understand and prepare, or am I using it to make a medical decision I should discuss with a clinician?

Patients should not need to become privacy lawyers to ask better health questions. The safe path should be obvious. Right now, it often is not.

So until the tools are clearer, use the cautious version.

Ask the question. Remove the identifiers. Avoid uploading photos unless you understand the privacy risk. Do not paste your whole chart when a general question will do. And if symptoms are urgent, worsening, or scary, call your doctor, urgent care, or emergency services instead of waiting for a chatbox to make you feel better.

AI can help you understand.

Do not give it more of your medical life than it needs.

Sources

1. HHS Office for Civil Rights. Guidance on HIPAA and Cloud Computing. https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html
2. American Medical Association. Augmented Intelligence Development, Deployment, and Use in Health Care. November 2024. https://www.ama-assn.org/system/files/ama-ai-principles.pdf
3. Office of the National Coordinator for Health Information Technology. HTI-1 Final Rule: Certification Program Updates, Algorithm Transparency, and Information Sharing. https://healthit.gov/regulations/hti-rules/hti-1-final-rule/
4. Coalition for Health AI. Blueprint for Trustworthy AI and Responsible AI Guidance. https://www.chai.org/workgroup/responsible-ai/blueprint-for-trustworthy-ai