OpenAI has unveiled a new plugin for ChatGPT that provides direct integration with the Apple Messages service (iMessage). According to the company, the tool is designed to make working with conversations more convenient: users can interact with their messages directly within a dialogue with the assistant, without switching between apps. The plugin runs on the Mac platform, where iMessage is the native messenger of the Apple ecosystem.
What the plugin can do: sorting, analysis, and editing
Among the key benefits listed by OpenAI is the ability to sort, analyze, and edit messages directly within a conversation with ChatGPT. The assistant can search for information in the "remote corners" of the chat history, delete messages on request, and, notably, write and send messages on the user's behalf. The plugin supports the Codex and ChatGPT Work services, which means the integration tools are available not only for personal use but also for professional activities — for example, for processing work chats and drafting replies in a business context.
Practical scenario: replying to all contacts in a single request
As a demonstration, OpenAI provides an example in which a user asks ChatGPT to suggest reply options for all the people they had been messaging the day before. In other words, the assistant doesn't just answer a single question — it goes through the entire previous day's correspondence and drafts personalized replies for each conversation. This scenario underscores the plugin's positioning as a "co-pilot" in the messenger: it takes on the routine work — reading, organizing, and preparing text — while leaving the final decision to send to the human.
Privacy: local operation and vague wording
The feature inevitably raises privacy questions, since the assistant gains access to the contents of personal and work chats. OpenAI assured that the plugin works locally on the user's device and "does not index all messages." However, the company did not clarify exactly what this wording means — it remains unclear what data is sent to the servers for generating responses and how long it is stored. This ambiguity is not a version conflict between sources, but it leaves room for criticism from privacy advocates.
Warnings: AI errors and mandatory confirmations
OpenAI explicitly warns that the AI makes mistakes, which in the context of messaging real people can be costly — from awkward phrasing to factual inaccuracies in work chats. Therefore, the company advises users to closely monitor ChatGPT's actions and enable persistent confirmations before sending messages on their behalf. This recommendation effectively acknowledges that fully autonomous message sending without human control carries an unacceptable risk, and shifts responsibility for the final text to the user.
