A new ChatGPT plugin promises to tame the black hole of old texts. Instead, it has turned the group chat into a test of who gets to decide when private conversation becomes AI-readable.

OpenAI rolled out its Apple Messages integration for ChatGPT on Mac last week, allowing users to search iMessage, SMS and RCS threads, summarize conversations, draft replies and send them. Setup requires explicit installation and permissions including AppleScript, Accessibility and Full Disk Access. OpenAI says the tool does not automatically index a Messages history: it reads a thread only when a user makes a request that requires it, and the material remains on the Mac by default.¹

That distinction has not calmed privacy critics. Security expert Paul Walsh argues that encryption is beside the point once a recipient’s computer can hand readable messages to another system. “Every single person I send a message to through iMessage will never know that I have a third party inside that application,” he said.¹ His concern is collective: one participant’s opt-in can expose years of exchanges written by people who never opted in.

Lookout CTO Dave Richardson offered a narrower but still cautionary view. Calling the spyware label “a little too strong,” he noted that the feature is off by default and permission-based; nevertheless, granting an AI access to messages means “you’re losing many of the benefits that end-to-end encryption has to offer.”¹ The stakes rise if a user chooses to save a ChatGPT conversation to the cloud, where any included message content follows the company’s ordinary retention rules.

The rollout also struck a cultural nerve. OpenAI’s developer-experience chief Romain Huet pitched the ability to ask who users talk to most, what they discuss and what follow-ups they missed.² Some users saw a productivity breakthrough; analyst Max Weinbach called it a “game changer,” while Yana Welinder welcomed a way to find things in iMessage’s “black hole.”²

Others drew a harder line: “If I find out that AI has been texting me on your behalf, I will never speak to you again,” one user wrote.² The dispute is no longer merely about a useful assistant. It is about whether the person holding the phone can unilaterally invite that assistant into everyone else’s conversation.