Quick Answer
ChatGPT Health is reasonably safe for general use, but connecting real medical records carries one tradeoff most coverage skips: once your records leave your provider, they lose HIPAA protection and are governed by OpenAI’s terms instead. OpenAI says it will not use the data to train models or target ads. Decide with that in mind.
Key Takeaways
- Connecting records to ChatGPT Health removes them from HIPAA, since that protection is legal, not technical, and does not travel with the data
- OpenAI states connected health data is not used to train its models or target ads, regardless of your other settings
- The feature is US-only, for logged-in users 18 and older, and it is read-only, so ChatGPT cannot change your records
- OpenAI’s privacy promises are real but voluntary, meaning they are set by policy and can change
- ChatGPT is explicitly not a diagnosis or treatment tool, and it does not replace a doctor
What is ChatGPT Health?
ChatGPT Health is a feature that lets you connect your medical records and Apple Health data so ChatGPT can answer questions using your actual health information. OpenAI rolled it out to US users on July 23, 2026, for logged-in adults 18 and older on web and iOS, across the Free, Go, Plus, and Pro plans.
Once connected, ChatGPT can draw on details like medications, lab results, recent visits, sleep, and activity. It can compare a new lab result against an old one, summarize what changed since your last appointment, or factor a health condition into an everyday question. You connect accounts through the Health section in the ChatGPT sidebar, and supported sources include Apple Health plus records from major systems such as Epic and Oracle Health.
The reason this matters is scale. OpenAI says more than 300 million people ask ChatGPT health questions each week, so this is one of its largest and most sensitive personal-data expansions. It also explains the redesign: an earlier, separate health space saw most users bypass it, with over 70% of health chats happening in ordinary conversations, so OpenAI moved the capability into general chat.
Does ChatGPT use your health data to train its models?
No, according to OpenAI’s own policy. The company states, in its Health support documentation, that connected medical records, Apple Health information, and any conversations that use that data are not used to train its foundation models or to target ads. That exclusion applies regardless of the general model-training setting on your account.
There is one boundary worth understanding, because it is easy to blur. The protection covers conversations that actually use your connected Health data. A health question you type in an ordinary chat, without invoking connected records, follows your standard model-training setting, which you control under Settings, then Data Controls. So the safest habit is to keep sensitive specifics inside the Health-connected flow rather than typing them into a general chat.
OpenAI also says all conversations are encrypted in transit and at rest, and that connected Health information receives additional encryption. Those are meaningful safeguards. The important qualifier, covered below, is that they are commitments of policy rather than legal guarantees.
Does connecting records to ChatGPT remove HIPAA protection?
Yes, and this is the single most important fact to understand before connecting anything. HIPAA is the US law that protects health information held by your doctor, hospital, or insurer. It protects the data because of who holds it, not because of what the data is.
When you copy your records into a consumer app like ChatGPT, they leave the custody of a HIPAA-covered entity. At that point the records are governed by OpenAI’s terms of service and applicable state privacy laws, not by HIPAA. This is a legal reality, not a flaw in ChatGPT, and no setting inside the app can restore the protection. As TechTimes reported, that protection is jurisdictional and does not travel with the data once you move it.
OpenAI itself is clear that ChatGPT Health is not built for clinical use and does not offer the business agreement that HIPAA compliance requires. Providers who need that are pointed to separate products. For an ordinary person, the takeaway is simple: connecting records is a deliberate trade of legal protection for convenience, so make it knowingly.
How safe are OpenAI’s privacy promises?
The promises are real, and they are also voluntary, and holding both ideas at once is the honest way to judge this. OpenAI’s commitments not to train on the data, not to serve ads from it, and to delete it within a stated window after you disconnect are genuine current policy.
The qualifier is that voluntary policy can change in a way a federal law cannot be changed by the company bound by it. A commitment set by terms of service today can be revised tomorrow, subject to notice and applicable law. That does not make the promises hollow. It means your protection rests on a company’s policy and its incentives rather than on a statute, which is a weaker and more changeable footing than HIPAA. Whether that tradeoff is acceptable is a personal judgment, not a settled fact, and reasonable people land in different places on it.
If you already track what you have exposed across services, this is one more account to log. Our guide to data privacy management tools covers ways to keep that inventory, and our overview of cybersecurity and data privacy covers the broader habits worth having.
Is ChatGPT Health accurate enough to trust?
ChatGPT Health is a capable tool that still makes mistakes, and it is explicitly not a substitute for a doctor. OpenAI is direct that ChatGPT is not intended for diagnosis or treatment and does not replace the judgment of a qualified professional.
On the numbers OpenAI has published, its latest models recognize when someone should seek emergency care more than 99% of the time in its own evaluations, while also avoiding unnecessary escalation at a similar rate (self-reported). Those figures come from the company’s internal testing, so treat them as vendor claims pending independent review. The models were also tested by physicians before release, per OpenAI.
The practical stance that follows is a middle one. ChatGPT Health can be genuinely useful for understanding a lab result, preparing questions for an appointment, or tracking a trend over time. It should not be the thing that decides whether you seek care. Stop here and contact a medical professional or emergency services, not a chatbot, if a symptom is serious or urgent. Use it to be a better-informed patient, not to replace the person treating you.
Who should and should not connect their records?
The decision splits by how sensitive your situation is and how much you value the convenience. There is no single right answer, so match it to your own case.
| Your situation | Reasonable approach |
|---|---|
| You want help understanding routine labs or tracking fitness | Connecting is a fair trade; the data is lower-sensitivity |
| You have a stigmatized or legally sensitive condition | Think hard before connecting; consider using it without records |
| You are in a custody, employment, or insurance dispute | Do not connect; records outside HIPAA can carry different exposure |
| You simply want general health answers | Use ChatGPT Health without connecting anything at all |
That last row is the option many people miss. The Health section works without connected records and can still answer health questions, so you can get the benefit of the improved health models without handing over your medical history. Connecting is a choice, not a requirement.
How do you delete your data or disconnect?
You stay in control, and disconnecting is straightforward, which is worth knowing before you connect so the decision feels reversible. ChatGPT Health is read-only, so it can view connected records but cannot alter them, and you decide each time whether a response may use your connected data.
By default, ChatGPT asks permission before using connected records in a response, and you can approve a single request or allow ongoing access. To pull back, disconnect the source in the Health section, and OpenAI states connected data is deleted within a stated window after you disconnect. You can also change your broader data settings under Settings, then Data Controls at any time. If you ever suspect a wider account problem rather than a Health-specific one, our guide on checking whether your email has been in a data breach is a sensible first step.
How does this compare to other AI health tools?
The pattern here is bigger than one company, and keeping it in view helps you make a consistent choice across services. Major AI providers are all moving into personal health, and each raises the same core question about where sensitive data ends up.
The specifics differ by product, but the principle is constant: any time you move health data from a HIPAA-covered provider into a consumer AI service, the legal protection changes, whichever company runs the service. Judge each on its stated policies, its track record, and how sensitive your data is, rather than on the brand name. For how the leading assistants compare more generally, our Claude versus ChatGPT versus Gemini overview is a useful starting point. If a shared health chat is a worry, our piece on shared AI chats appearing in search explains a related exposure.
FAQ
Is ChatGPT Health safe to use?
It is reasonably safe for general questions, and OpenAI applies encryption and says it will not train on or sell your health data. The main tradeoff is that connecting real medical records removes them from HIPAA protection, leaving them under OpenAI’s policies rather than federal law.
Does connecting my records to ChatGPT remove HIPAA protection?
Yes, because HIPAA protects health data based on who holds it, so once records leave your provider for a consumer app they fall under OpenAI’s terms and state privacy laws instead. No setting inside ChatGPT can restore HIPAA coverage, because the protection is legal rather than technical.
Does OpenAI use ChatGPT Health data to train its AI?
No, per OpenAI’s stated policy: connected medical records, Apple Health data, and conversations that use them are excluded from model training and ad targeting, regardless of your other settings. Health questions typed into ordinary chats without connected data follow your standard model-training setting instead.
Can ChatGPT Health diagnose or treat me?
No, because OpenAI states ChatGPT is not intended for diagnosis or treatment and does not replace a qualified medical professional. It can help you understand results and prepare for appointments, but for anything serious or urgent, contact a doctor or emergency services instead.
How do I delete my data from ChatGPT Health?
Open the Health section in the ChatGPT sidebar and disconnect the source you added, and OpenAI states connected data is deleted within a stated window afterward. You can also adjust broader data settings under Settings, then Data Controls at any time.