Story
August 22, 2026

ChatGPT Can Now Text From Your Mac, but the Group Chat Draws a Line

OpenAI’s Apple Messages plug-in promises faster searching, drafting and follow-ups on Mac. But as the tool reaches into intimate conversations, users are split between workplace convenience and fears over privacy, consent and sincerity.

OpenAI has brought ChatGPT into Apple Messages on the Mac, turning the assistant from a writing aide into a potential participant in the most personal corner of the internet: the group chat. The promise is convenience; the immediate backlash is about who, exactly, is doing the talking.

The company unveiled the plug-in Thursday for ChatGPT Work and Codex users on desktop. It can search message histories, help users catch up on conversations, draft replies and send them; OpenAI has also suggested using it to find birthday or spam messages for deletion. In a product demonstration, the system was shown proposing follow-ups based on messages received the previous day.

OpenAI’s pitch is that the feature makes an unruly inbox useful again. Romain Huet, the company’s head of developer experience, suggested asking the tool “who you talk to most, what you talk about, or what you forgot to follow up on!” Supporters see a genuine productivity gain: analyst Max Weinbach called it a “game changer” for text-heavy work, while Yana Welinder said it could finally help her retrieve information from the “black hole of iMessage.”

But the launch landed differently among people who treat texting as a proof of personal effort. One user’s warning was blunt: “If I find out that AI has been texting me on your behalf, I will never speak to you again.” Another said AI-written emails or texts made a sender “effectively dead to me,” arguing that a relationship without effort was not worth maintaining.

Privacy is the other fault line. OpenAI says the plug-in runs locally and “doesn’t create an index of all someone’s messages,” though the practical limits of that assurance remain unclear. The company says installation and access are explicit, and that ChatGPT reads a message only when prompted. It also warns users not to enable persistent approval, which removes their final chance to review a message before it is sent in their name.

That leaves the new feature caught between two visions of messaging: a searchable, automatable workflow for one side, and a stubbornly human social space for the other.