ChatGPT plugin reportedly reads and replies to iMessages on Mac for Work and Codex users

OpenAI’s plugin bridges ChatGPT with Apple’s Messages app, allowing users to delegate tasks like searching conversations or drafting replies. This marks a shift from browser-based interactions to direct control over native macOS applications. The feature targets enterprise users, suggesting OpenAI is prioritizing productivity workflows over consumer adoption. However, the requirement for Full Disk Access and contact permissions creates a high barrier to entry, limiting its appeal to users comfortable with broad system-level permissions.

The integration’s technical implementation remains unclear. OpenAI has not disclosed whether it collaborated with Apple or reverse-engineered iMessage access. Apple’s past actions, such as disabling Beeper Mini’s iMessage integration, indicate a strict stance against unauthorized access to its services. The ongoing legal dispute between Apple and OpenAI further complicates the plugin’s future, as Apple could revoke access or impose restrictions at any time. Engineers evaluating this tool must weigh its utility against the risk of sudden deprecation.

Privacy and security implications dominate this release. Granting ChatGPT access to message history and contacts exposes sensitive data to OpenAI’s infrastructure, even if the feature is opt-in. The plugin’s reliance on on-device data processing may mitigate some risks, but the requirement for Full Disk Access could expose unrelated files if misconfigured. Organizations adopting this tool will need to update their data handling policies and educate users on the trade-offs between convenience and exposure.

For developers, this plugin signals a growing trend of AI agents interacting with native applications. The ability to automate messaging tasks could streamline workflows in customer support, scheduling, or internal communications. However, the lack of transparency around Apple’s involvement and the plugin’s enterprise-only availability suggest it is an experimental feature rather than a stable integration. Teams should treat it as a proof of concept until OpenAI provides clearer documentation and Apple’s stance becomes more definitive.