OpenAI just gave ChatGPT access to your text messages. Not a summary feature, not a read-only preview — actual read, draft, and send access to your Apple Messages inbox. According to TechCrunch, OpenAI has launched an Apple Messages plugin for ChatGPT that lets users connect their Messages inbox directly to the chatbot. It’s one of the more personal integrations any major AI company has shipped, and it raises real questions about where the line is between helpful and overreaching.
The plugin works locally on the user’s machine, which is the detail OpenAI is leading with to address obvious privacy concerns. The company told Bloomberg it doesn’t create a full index of your messages. That’s reassuring in theory, but the specifics are vague enough that it’s worth waiting for a clearer technical explanation before treating it as a green light. TechCrunch has asked OpenAI for more details on exactly what data is accessed and when.
What the plugin actually does is fairly broad. Users can ask ChatGPT to sort or analyze their message history, draft replies, search for information buried across old conversations, and delete messages entirely. A promotional clip shows someone asking ChatGPT to suggest follow-up messages based on what their contacts sent the previous day. That’s a narrow use case, but it hints at how much behavioral and relational data the model would need to touch to do that well.
- Sort and analyze existing messages
- Draft and send messages on your behalf
- Search across message history
- Delete messages
- Suggest context-aware follow-ups based on prior conversations
The plugin also connects with Codex and ChatGPT Work, which opens it up to professional use. That’s a meaningful detail. If teams start using this in work contexts, where Messages threads often contain client information, project details, or internal decisions, the privacy stakes get considerably higher than they are for personal texting.
OpenAI is clearly aware of the risks that come with send-on-your-behalf functionality. The company actively discourages users from enabling persistent approval, which would let ChatGPT send messages without confirmation. The warning is direct: doing so removes your final chance to review a message before ChatGPT sends it as you. That’s an honest caveat, and it suggests OpenAI knows this is the feature most likely to go wrong.
This launch fits into a broader push by OpenAI to make ChatGPT genuinely agentic rather than just conversational. Google has been moving in a similar direction with Gemini’s integration into Gmail and Google Messages. Apple Intelligence, which sits natively on-device, also handles some message summarization and reply suggestions, but without the same level of third-party AI access. OpenAI is essentially asking users to extend more trust to ChatGPT than Apple has claimed for itself, which is a notable positioning choice.
For developers and founders building on top of ChatGPT, this signals that OpenAI sees plugins with real-world write access as a core part of the product’s future. But for everyday users, the question is simpler: how much do you trust any AI company with the full text of your personal conversations? That answer should drive whether this plugin ends up being useful or something you install once and quietly disconnect.
