ChatGPT Health adds privacy costs Siri AI can limit

OpenAI’s new Health feature gives ChatGPT access to Apple Health data and medical records, adding useful analysis at the cost of sharing sensitive information with another company.

The feature is available to logged-in users 18 and older across Free, Go, Plus, and Pro plans. OpenAI says ChatGPT can compare laboratory results over time, summarize changes since a medical appointment, and identify patterns involving sleep, exercise, and daily activity.

Users can connect Apple Health, medical records from supported U.S. hospital systems, One Medical, and Function Health. ChatGPT can also receive information that fitness, nutrition, and wearable apps make available through Apple Health, although some proprietary scores may not transfer.

OpenAI says more than 300 million people turn to ChatGPT with health-related questions each week. Connecting records could make it easier to understand medical terminology, prepare questions for a doctor, or track changes without repeatedly uploading the same documents.

Health in ChatGPT requires users to sync potentially extensive medical and wellness information with OpenAI. Users who don’t connect the feature avoid adding OpenAI to the services handling that data.

OpenAI promises safeguards, but the data still leaves Apple

OpenAI says connected medical records, Apple Health information, and related conversations won’t train foundation models or target ads. It further claims that all ChatGPT conversations are encrypted in transit and at rest, with added protections for Health data.

ChatGPT asks for permission before using connected health information by default. Users can approve access for one request or allow ChatGPT to use it automatically in future conversations.

Permission controls don’t eliminate the underlying data transfer. ChatGPT can analyze health information only after users make it available to OpenAI.

Percentage of reviews earning high scores by evaluation category. Image credit: OpenAI

OpenAI deletes synced health data within 30 days after disconnection, while information in past chats remains until users delete those conversations. ChatGPT can also create memories from Health conversations, and sharing health details through another service may require confirmation.

Siri AI offers a first-party alternative with fewer data transfers

Apple’s Siri AI in iOS 27 uses a first-party architecture for supported requests involving personal context. Siri AI can search across apps for user data, understand information shown onscreen, and perform actions through Apple’s systemwide app frameworks.

The company processes many requests on the device and uses Private Cloud Compute when a request needs a larger model.

Private Cloud Compute sends only the data needed to process a request to Apple Silicon servers. Apple can’t access or retain that data, and researchers can inspect the software used to process it.

The key difference is how Apple and OpenAI process the data. For supported health tasks, Apple can process requests on-device or through Private Cloud Compute without a persistent connection to OpenAI.

Processing supported health requests on-device or through Private Cloud Compute avoids adding OpenAI as another service handling the data. OpenAI’s safeguards reduce some risks, but avoiding the extra connection minimizes data sharing more effectively.

Apple and OpenAI remain partners while fighting in court

The Health rollout also comes amid a legal conflict between Apple and OpenAI. Apple continues to offer optional ChatGPT integration through Apple Intelligence.

Connecting Apple Health to ChatGPT. Image credit: OpenAI

Apple sued OpenAI and two former Apple employees on July 10 over alleged trade secret theft tied to OpenAI’s consumer hardware work. The lawsuit names Tang Yew Tan and Chang Liu.

According to the complaint, both men improperly accessed, kept, or shared confidential information connected to their departures from Apple and later work for OpenAI.

OpenAI has denied wanting or using competitors’ trade secrets. The allegations haven’t been proven, and the dispute is separate from the privacy and security of Health in ChatGPT.

Health in ChatGPT still requires careful verification

Hundreds of physicians helped OpenAI build realistic health scenarios for testing ChatGPT. The evaluations measure accuracy, safety, communication, context awareness, completeness, and whether ChatGPT recognizes when professional care may be needed.

GPT-5.6 outperformed GPT-5.5 on OpenAI’s HealthBench Professional evaluation, though ChatGPT can still make mistakes. Users should verify important information with a qualified health care provider.

Connected records may also be incomplete or outdated. OpenAI notes that a medication can remain listed after someone stops taking it, making it important to check ChatGPT’s information against the original source.

ChatGPT can make records easier to understand, identify changes worth discussing, and help someone prepare for an appointment. Its output shouldn’t be the sole basis for a diagnosis, medication change, or decision to avoid professional care.

Health in ChatGPT organizes scattered medical data. Apple users must decide whether that convenience is worth sharing their health information with OpenAI.