
OpenAI launches a ChatGPT plugin for Apple Messages that can basically text on your behalf
OpenAI has released a new ChatGPT plugin for Mac that gives the AI direct access to your Apple Messages conversations. Available through ChatGPT Work and Codex on Apple Silicon Macs, the plugin can search through message history, summarize and analyze conversations, draft replies, suggest follow-ups, and even send or delete messages on the user’s behalf. For now, the integration is limited to macOS and is not available through ChatGPT on iPhone or iOS.
That level of access has naturally raised privacy concerns, especially given how much personal information can live inside Messages and Apple’s traditionally strict approach to letting third party apps access users’ personal data. The launch also comes amid an ongoing legal dispute between Apple and OpenAI over alleged trade secret theft. To use the integration, users need to connect Messages to ChatGPT and grant several macOS permissions, including Full Disk Access, contact access, and system automation. The plugin then uses tools such as AppleScript and Accessibility to interact directly with the Messages app.
OpenAI says the plugin runs locally, doesn't create a full index of a user’s message history, and only accesses message content when specifically requested. Message data is stored locally by default, while anything saved to the cloud follows OpenAI’s standard retention policies. The company also recommends reviewing actions before messages are sent, warning that persistent approval removes the user’s final chance to check what ChatGPT is about to send in their name. Personally, I can see this being handy for dealing with companies or other bots, but for actual conversations with friends and family, though? Yeah, probably not.




Comments
Who even wants this? Like Human Connection is a desirable and healthy thing.
Interacting by voice is a real human connection. Texting barely qualifies but still better than auto reply.
I may be mistaken, but news on AlternativeTo seems to be increasingly editorialized lately. ("The whole situation couldn’t be more ironic"; "If you still want to try it anyways"; "Yeah, probably not") Serious privacy concerns are absolutely worth drawing attention to, but it's worth considering that subtlety might express them better.
Hey, thanks for the feedback! I didn’t mean for it to come across that way, but also wanted to add a bit of personal opinion so it wouldn’t read like a simple summary. I’ve adjusted the wording in a few places so it fits a bit better though 🙂
Come on, you didn't have to change anything. I just meant it as something to consider. Personal opinion is appreciated, so long as you point out that that's what it is— which you did.
Oh my god. Instead of chatting with your friends, you can have your AI bot chat with their AI bot. Now you don't have to deal with the annoyance of talking to the people in your life!
Devils advocate. Set AI bot only for people who are not in your contact list. That could be useful to avoid spam. Or maybe you are skipping work, and the AI bot answering text creates impression that you are working from home. But avoiding to talk with real contacts. That is self defeating.