Apple Health AI Study With OpenEvidence: How Your Data Is Handled
Apple's Health AI Usage Study app, launched on 15 November 2026 with OpenEvidence, lets US iPhone users aged 21 and over ask health questions and get cited answers. Apple says data is anonymized before going to a third-party cloud provider, is not used to train AI, and that OpenEvidence deletes queries and answers after 24 hours. Sharing Health app data is optional.
On this page
What Apple launched
Apple released the Apple Health AI Usage Study app in the US App Store on 15 November, built with OpenEvidence, a medical AI company whose tool is used by doctors to search clinical research. Apple announced it at the HLTH conference in Las Vegas.
The app answers basic and complex health questions using peer-reviewed sources; TechCrunch names NEJM, JAMA, and Nature among them. Answers come with linked citations, and Apple uses its own technology to turn the research into plain language. Participants rate answers and complete surveys, and Apple says the study will measure trust, comprehension, helpfulness, and whether AI helps people prepare for doctor visits.
This is a research study, not a product launch. TechCrunch notes it is unclear whether Apple will bring AI health answers to the Health app or Siri.
Who can join
Apple wants up to 200,000 participants, and TechCrunch reports the study will run for more than 12 months. It is open to US adults aged 21 and over on iOS 26 or later. Device requirements differ in coverage: ABC News and Chief Healthcare Executive say an iPhone 15 Pro or later, while TechCrunch says an iPhone 15 or later with Apple Intelligence enabled. Since Apple Intelligence needs a 15 Pro or newer, the Pro requirement is the safer reading.
Participants go through an informed consent process inside the app. Chief Healthcare Executive points out that the age and device rules may skew who is studied toward people who can afford recent iPhones.
How your health data is handled
You can choose whether to share Health app data such as resting heart rate, activity, sleep, or lab results to get more personal answers. Apple's stated protections are:
- Data is anonymized before it is sent to a secure third-party cloud provider, which Apple has not named.
- Study data will not be used to train AI models.
- OpenEvidence deletes question and answer data after 24 hours and will not train its model on the queries, according to Apple.
ABC News describes the step differently, saying protections are designed to remove identifying information before questions reach OpenEvidence. Removing names and identifiers is not the same as true anonymization, and health questions themselves can be revealing. The app also includes safeguards that flag potentially dangerous situations and point users to professional help, and Apple stresses it is not live monitoring and does not replace a clinician.
The privacy trade-off
Apple usually markets health and AI features around on-device processing. This study works differently: your questions, and any Health data you opt to include, leave the phone for a cloud provider and a third-party AI company. None of the coverage says any of the answering happens on the device. All of the protections above are Apple's own statements, and no independent audit has been published.
If you join, treat it like any cloud health chatbot. Share only the Health categories a question needs, avoid names and other identifying details in your prompts, and remember you can decline to share Health data at all. For questions you would rather keep entirely private, an AI model that runs locally on your own device keeps the text from leaving it.