ChatGPT can now read your Apple iMessages — and critics are calling it a 'betrayal' of trust

The plugin, announced Aug. 20, is touted by ChatGPT as a helper that “can read and search iMessage, SMS, and RCS chats” and “send messages on your behalf through the Messages app.” The platform adds that messages are only sent after user approval but that a function exists to allow ChatGPT to send messages automatically “without another send approval.”

While some on social media applauded the new tech, a large number of commenters warned about the implications of allowing an AI chatbot to access your personal messages.

“Isn’t this like a privacy nuclear bomb though?” one asked on a ChatGPT X post about the new feature. Another questioned how users can opt out of having their messages scraped by ChatGPT if someone else in their private chats opts in.

And developer Steve Moraco added his own stark rebuke to the privacy-focused iPhone maker: “Total architecture abandonment and user trust betrayal on Apple’s part.”

A privacy promise meets an AI chatbot

OpenAI, which owns ChatGPT, reassured TechCrunch that the new plugin runs locally on someone’s device, doesn’t fully index messages, can’t run without user permissions and that messages are not saved to the company’s servers. They added that if conversations are saved in the cloud, the data retention policy is reflective of the rest of the content saved there.

Neither OpenAI nor Apple immediately replied to Moneywise’s request for comment.

Moraco, however, pointed to the fact that iMessage is “quantum encrypted” to protect information and that, if someone he previously messaged opts in to the plugin, “copies of my messages from years ago on [anyone’s] laptop … can now be fetched directly by ChatGPT without my permission or even knowledge.”

Online safety and privacy expert Paul Walsh also warned against using the new plugin, alleging on Substack that it is akin to “installing spyware that can be exploited by governments, intelligence agencies, law enforcement, internal threat actors at OpenAI and partners.”