What stays local, what can sync, and what reaches AI.
HeyCC has several distinct data paths. App storage can remain on the device, supported data can use optional iCloud sync, and relevant request content can be sent to the AI model service you configure. Apple system permissions and HeyCC action rules form separate control layers.
The short answer
- On device: app configuration and supported local data.
- Through iCloud: only supported data when the related setting is enabled.
- To a model service: relevant messages and attachments required for an AI request.
- Through system access: Calendar, Reminders, microphone, files and similar features remain subject to platform permission and the requested action.
Four controls, not one privacy switch
HeyCC stores app data on the device. Local storage describes where app data lives; it does not by itself describe where an AI request is processed.
When the related setting is enabled, supported data may sync through the user’s own iCloud. Sync coverage varies by data type, platform and version, so it should not be described as an automatic backup of everything.
Relevant messages, instructions and attachments needed to answer a request can be sent to the supported model endpoint selected by the user. Provider access, terms, retention, privacy and fees apply there.
Using Calendar, Reminders, files, microphone or another system feature depends on the operation and permission state. Choosing a model does not grant Apple permission, and Apple permission does not approve every HeyCC action.
A concrete request path
Example: you ask a bot to turn a voice note into a task list and add approved items to Reminders. The recording or transcript begins in the app workflow → the relevant request content can go to the selected model provider for processing → the response returns to HeyCC → creating reminders remains a separate system action subject to permission and review. If supported sync is enabled, eligible app data may also follow its own iCloud path.
This separation matters because “stored locally,” “synced with iCloud,” “processed by an AI provider,” and “allowed to change another app” are different facts. None should be used as shorthand for the others.
Files and attachments
A file can have more than one relevant destination. The original may remain in app or system storage, content required for the request may be sent to the selected model provider, and a generated result may be written to a destination chosen by the user. Before approving a file workflow, check:
- which file or excerpt is actually needed;
- which model service is selected for the request;
- where the output will be written and whether it can overwrite an existing file;
- whether the finished format opens correctly before sharing it;
- whether the provider or destination has a separate retention policy.
Bot roles help organization, not isolation
Giving a bot a name, appearance and role can keep work and personal workflows understandable. It is not an automatic security boundary. Actual access depends on the data placed in the conversation, enabled tools, provider configuration, system permissions and action rules. A GroupBot or role-based bot should not be described as isolated unless that boundary has been separately verified.
Before using sensitive information
- Confirm the selected provider and review its current terms and privacy policy.
- Reduce the request to the minimum messages, files and fields required.
- Check whether relevant iCloud sync is enabled and what the installed version supports.
- Review Calendar, Reminders, microphone, files and other system permissions involved.
- Inspect the proposed action and destination before allowing a write, send, upload or overwrite.
- Use non-sensitive test data when the full path has not been verified.
This guide is a product explanation, not a substitute for the current privacy policy, provider disclosures or in-app permission screens.
HeyCC privacy and data-flow FAQ
Does HeyCC run every AI request entirely on the device?
No. Relevant messages and attachments needed for an AI request are sent to the supported model service selected by the user. Device-side app storage and tools do not mean that model processing is offline.
Does all HeyCC data sync through iCloud?
No. Sync is selective, depends on supported data types, settings, platform and version, and some data remains device-local.
What can reach the selected AI provider?
Relevant messages, instructions and attachments needed to process a request can be sent to the configured model endpoint. The provider’s terms, privacy policy, retention rules and fees apply.
Does choosing a model also approve Calendar, Reminders or file actions?
No. Model configuration, Apple system permission and HeyCC action rules are separate controls. One does not automatically grant the others.
Are separate personal bots automatic security boundaries?
No. Separate names and roles help organize work, but actual access depends on configuration, connected tools, permissions and the data included in each request.
What should I check before using sensitive information?
Check the selected provider, exact request content, enabled sync, system permissions, destination files and current policies. Use non-sensitive test data when a path is not yet verified.