ChatGPT gets Apple Messages plug-in that can draft and send texts
OpenAI has launched an Apple Messages plug-in for ChatGPT that lets users read, search, draft, and send iMessages directly through the chatbot.
👁 116 reads
OpenAI launched an Apple Messages plug-in for ChatGPT on August 20, 2026, giving users the ability to connect their Messages inbox to the chatbot and have it draft, send, delete, or search through texts on their behalf.
- The plug-in connects a user's Apple Messages inbox to ChatGPT, enabling sorting, analysis, drafting, and sending of messages
- It also works with Codex and ChatGPT Work for professional use cases
- OpenAI says the plug-in runs locally and does not create an index of all messages, though full details remain unclear
According to the company, the plug-in supports a range of actions beyond simple drafting. Users can ask ChatGPT to sort and analyze their message history, suggest follow-up replies based on conversations received the previous day, or surface information buried deep in older threads. The integration also extends to Codex and ChatGPT Work, making it available for professional as well as personal use.
Privacy is an immediate concern. OpenAI told Bloomberg that the plug-in runs locally on a user’s machine and “doesn’t create an index of all someone’s messages,” though the company has not yet provided full technical details on what that means in practice. TechCrunch reported it had reached out to OpenAI for clarification.
On the question of autonomous sending, OpenAI takes a cautious public stance. The company actively discourages enabling persistent approval, warning that doing so “removes your final chance to review a message before ChatGPT sends it as you.” The recommended approach keeps the user in the loop before any message is transmitted.
Practitioner takeaway: For anyone who already uses ChatGPT heavily for communication tasks - writing follow-ups, summarising threads, managing a high-volume inbox - this plug-in brings that workflow directly into Apple Messages without copy-pasting. The practical risk is the auto-send path: the plug-in is capable of sending messages as the user, and OpenAI’s own warning makes clear that turning off the review step means losing a meaningful safeguard. Until the local-processing claim is independently verified, connecting a personal or business Messages inbox is a decision that warrants reading whatever privacy documentation OpenAI publishes alongside the launch.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →
More on this from QUANTSLANT
ai tools·Grok Chatbot Sends Gibberish Responses to Users on Grok.com