ChatGPT can now read and send your Apple Messages

If you are drowning in unread texts or constantly digging through weeks of chat history to find a single detail, OpenAI wants to handle that chore for you.

The company just launched an official Apple Messages plugin for ChatGPT on Mac, allowing users to connect their Messages inbox directly with the chatbot. The feature is available immediately on the desktop versions of ChatGPT Work and Codex, and users can install it by navigating to Plugins and selecting Public.

What the plugin actually does

Once set up, ChatGPT can search your message history, catch you up on missed conversations, and even draft and send replies on your behalf.

In demonstrations, the plugin handles several everyday tasks quite naturally. You can ask ChatGPT to review messages from yesterday and list all required follow-ups, or have it check your calendar to suggest meeting times based on inbound texts. It can also spot birthdays buried in conversations, retrieve deeply hidden information, and flag suspected spam messages for deletion.

Because it integrates with both ChatGPT Work and Codex, OpenAI is pitching this tool for professional productivity as well as personal catch-up.

Local processing with a big permission catch

Letting an AI read your personal texts naturally triggers privacy alarm bells, and OpenAI quickly offered details on how the architecture works.

According to the company, the plugin runs locally on your Mac, and ChatGPT does not build a full index of your messages. It only reads your messages when you give it a specific request to do so. In addition, desktop conversations are stored locally on your machine by default, and text content is not saved to OpenAI servers unless you manually choose to sync those chats to the cloud.

However, there is a catch: setting up the plugin requires granting ChatGPT Full Disk Access on macOS. This gives the application extensive read and write capabilities across your local drive.

OpenAI also explicitly discouraged users from enabling persistent approvals. The company warned that turning off manual confirmations removes your final chance to review a message before ChatGPT sends it out under your name.

Why it matters

We see this as a significant move for OpenAI, shifting ChatGPT from a browser-bound text generator into an active personal assistant embedded in your daily operating system.

For power users and busy professionals on Mac, the utility is obvious. Having an AI that can extract action items and draft contextual replies can save serious time every day. Still, granting Full Disk Access is a massive permission to hand over, and the bar for trust here is extraordinarily high.

OpenAI's decision to keep processing and storage local is a smart move that softens the privacy hit. But letting an AI hit the send button remains risky. While letting ChatGPT summarize your texts is a no-brainer, we think you should definitely keep manual approval turned on before letting any model speak for you.