ChatGPT Can Text Your Ex Now

OpenAI's new plugin drafts, sends, and deletes your iMessages for you. It's also asking to be let into every conversation you've ever had.

Welcome Back,

Ever wanted someone else to do your texting for you? OpenAI just launched an Apple Messages plugin for ChatGPT, letting interested users connect their Messages inbox directly with the chatbot, according to TechCrunch. Once connected, ChatGPT can sort, analyze, or edit your messages, and the plugin also works with Codex and ChatGPT Work, meaning it's built for professional use as much as personal texting.

A promotional clip advertising the new plugin shows a user asking the chatbot to suggest follow-up messages to their contacts based on what came in the day before. Beyond that, you can also ask ChatGPT to delete messages for you, draft and send messages on your behalf, or search for information buried deep in your message history.

Today we look at exactly what the plugin lets ChatGPT do, the specific privacy claim OpenAI made to Bloomberg and why it's still short on detail, the one safety setting OpenAI explicitly warns against turning on, and what this launch says about where AI assistants are actually headed next.

💬 What the Plugin Actually Lets ChatGPT Do

Once connected, ChatGPT gains a genuinely broad set of capabilities over your Messages inbox. It can sort and analyze your conversations, edit messages, draft and send new ones on your behalf, delete messages, and search for specific information buried somewhere in your message history, the kind of thing that would normally take you scrolling back weeks or months to find.

That's a meaningfully different category of feature than most chatbot integrations. This isn't ChatGPT reading a document you upload or answering a question, it's ChatGPT actively operating inside one of the most personal communication channels most people have, with the ability to both read what's already there and generate new messages that go out under your name.

That professional angle is worth noting on its own. Extending this into work contexts through Codex and ChatGPT Work suggests OpenAI sees this less as a novelty feature and more as a serious step toward ChatGPT becoming a genuine communications assistant, not just a chat window you copy answers out of.

The plugin can draft, edit, and send messages on your behalf across your existing Apple Messages conversations.

🔒 "Doesn't Create an Index," But What Does That Actually Mean?

As with most things related to AI, this feature raises real privacy questions, and OpenAI's initial answer leaves a fair amount unexplained. OpenAI told Bloomberg that the plugin runs locally on a user's machine and that it "doesn't create an index of all someone's messages."

That gap between a reassuring-sounding phrase and a fully specified privacy architecture is exactly the kind of detail worth watching for as OpenAI, or TechCrunch's follow-up reporting, fills in the blanks. Until then, "doesn't create an index" is a real but genuinely partial answer to the bigger question of where your message data actually goes.

⚠️ The One Setting OpenAI Tells You Not to Turn On

When it comes to actually sending messages, OpenAI encourages users to keep an eye on what ChatGPT is doing, and specifically discourages turning on persistent approval, the setting that would let ChatGPT keep sending messages on your behalf without asking each time.

OpenAI's own documentation is direct about why, warning that persistent approval "removes your final chance to review a message before ChatGPT sends it as you." That's a genuinely blunt admission, framed entirely in terms of losing your own oversight, not in terms of any technical guarantee about accuracy or appropriateness.

OpenAI's own guidance frames message review as your last safety check before something goes out under your name.

📈 Part of a Bigger Pattern: AI Reaching Into Real Life

This launch fits a pattern that's been accelerating across the AI industry lately, chatbots moving beyond answering questions in a browser tab and gaining direct read-and-write access to the apps people already use daily, email, calendars, messaging, and now iMessage specifically.

That shift is genuinely useful in a lot of daily scenarios, faster replies, easier organization of a cluttered inbox, quick retrieval of an old address or plan buried in a months-old thread. It also raises the stakes meaningfully, because a mistake now isn't a wrong answer you notice and ignore, it's potentially a message sent to a real person before you ever see it.

Apple's own Messages ecosystem has historically been a relatively closed, tightly controlled space. A third-party AI assistant getting this level of access is a meaningful shift in how open that ecosystem is becoming, and worth watching for how Apple itself responds, whether through its own competing features or additional privacy safeguards.

⚖️ What's Genuinely Useful vs. What's Still Unclear

It's worth separating what this plugin clearly does well from what remains genuinely uncertain about how it works underneath the marketing.

None of that means the plugin is a bad idea, sorting a cluttered inbox and surfacing an old buried detail are both genuinely useful. It just means the sensible approach right now is caution, review every message before it sends, and pay attention to what OpenAI eventually clarifies about exactly how "local" this processing really is.

Reviewing every AI-drafted message before it sends is currently the main safeguard users actually control.

🧠 AI Spotlight Analysis

This is a small feature announcement with genuinely large implications if it becomes widely adopted. Text messaging is one of the last remaining communication channels that's felt distinctly human and unmediated, unlike email, which has been filtered through spam filters and smart replies for years. Handing that channel over to an AI assistant, even partially, is a real shift in how personal AI's role in daily life is becoming.

OpenAI's own cautious framing around persistent approval suggests the company itself understands the stakes here, this isn't being marketed as a fully autonomous feature, it's being launched with an explicit nudge toward keeping a human in the loop. That's the right instinct, and it's a genuinely different posture than simply shipping the most capable, most autonomous version of the feature and letting users figure out the risks themselves.

That single line is the most important sentence in the whole announcement. It's a genuine, direct acknowledgment that the safest version of this feature is the one where a human still checks the AI's work before it reaches another person, and it's worth taking that advice at face value rather than optimizing for convenience over review.

🔗 Sources and Further Reading

Thanks for reading AI Spotlight.

Our mission is simple: deliver clear, trustworthy, and actionable AI insights that help professionals stay ahead without the hype.