OpenAI Links ChatGPT Directly to Apple Messages on Mac With Deep System Access

OpenAI has released a new integration that connects ChatGPT directly to Apple Messages on macOS, enabling the artificial intelligence assistant to read past text conversations, draft responses, and send messages on behalf of users.

The feature is currently restricted to subscribers on the ChatGPT Work and Codex tiers. Demonstration materials shared by OpenAI show the desktop chatbot scanning for unread communications from the previous day and generating outgoing replies based on conversation context.

Users must grant ChatGPT permission to access local message databases, authorize contact name reading, and enable system automation tools, manually bypassing multiple layers of native macOS security controls to activate the functionality.

Users must also navigate to System Settings to grant the application Full Disk Access, an elevated security permission that provides unrestricted access to nearly all stored user data on the computer.

Addressing data handling practices, an OpenAI spokesperson stated that ChatGPT accesses message threads only when specifically prompted by the user to analyze context.

The tool does not build a permanent index of a user’s text history, and the company also confirmed that outgoing texts require manual user confirmation before transmission, stating that the tool was engineered to preserve user control over communications.

The release arrives amid a strict precedent set by Apple regarding unauthorized third-party integrations with its proprietary messaging infrastructure. In 2024, Apple repeatedly blocked reverse-engineered access attempts by the Beeper Mini chat service until the developers ceased workarounds.

The rollout also coincides with active litigation between the two companies, following a lawsuit filed by Apple in July that accused OpenAI of trade secret theft linked to the recruitment of former employees.