ChatGPT can now Summarise and Send Your iMessages
Last update: August 21, 2026
Disclaimer: This website may contain affiliate links, which means we may earn a commission if you click on the link and make a purchase. We only recommend products or services that we personally use and believe will add value to our readers. Your support is appreciated!

Imagine telling an AI, "find that address Mum sent me last Tuesday and reply saying I'm on my way." You can now do that — because ChatGPT can officially slide into your Apple Messages.
As first reported by cbinews.tv, OpenAI quietly rolled out a new Apple Messages plugin on Thursday, 20th August, and it's already got Apple-watchers talking.
Yes, you read that right. The ChatGPT app for Mac can now control your Messages app. We’re talking read, write, send, search, summarise — the whole lot.
So, what’s actually happened?
Until now, if you wanted AI help with your texts on a Mac, you were stuck with Siri — and let's be honest, Siri wasn't exactly setting the world alight.
OpenAI has changed the game. If you install the new Apple Messages plugin from ChatGPT's Plugins > Public section, you can ask things like:
"Catch up on my chat with David"
"Summarise my unread messages from today"
"Check my calendar and reply to Sarah with times I'm free next week"
It's available in ChatGPT Work and Codex on desktop for all plans, according to OpenAI.
Behind the scenes, it's not some dodgy hack. OpenAI says it runs locally on your Mac and piggybacks on existing macOS tools like AppleScript and Accessibility features.
But — and it's a big but — you have to give it some pretty serious access. We're talking Full Disk Access in System Settings, plus permissions for Contacts and automation. You'll get several opt-in screens in both ChatGPT and macOS before anything happens.
OpenAI is keen to stress: you have to consent, and it doesn't create a permanent searchable index of all your messages.
Why this is a big deal for Apple
This touches a nerve, because privacy is Apple's brand.
Since Tim Cook's famous 2015 open letter and the launch of App Tracking Transparency in 2021, Apple has marketed iMessage as end-to-end encrypted, private-by-design. Over 1.4 billion active Apple devices rely on that promise. Apple even famously went to war with the FBI in 2016 over unlocking an iPhone, and more recently shut down Beeper Mini in late 2023 for trying to reverse-engineer iMessage for Android.
So, when a third-party AI — especially one as powerful as ChatGPT — can suddenly read your private chats, even locally, it raises eyebrows. Apple hasn't commented yet, but you can bet they're watching.
There's another layer here too. Apple is building its own version of this. Its much-delayed, AI-supercharged Siri — expected with Apple Intelligence — is supposed to do exactly this: search through your messages, emails and apps to give you personal context. OpenAI just got there first on Apple's own turf.
This isn't OpenAI's first deep dive into your device either. The company recently added support for analysing data from Apple Health on iPhone and iPad, and ChatGPT for Mac already integrates with Xcode, Notes and Terminal. But Messages is different. It's personal.
What about safety?
Here is where it gets a bit more serious, especially for us in Nigeria.
You might remember that in December 2025, NITDA's Computer Emergency Readiness and Response Team (CERRT.NG) issued an advisory about vulnerabilities in GPT-4o and GPT-5. Researchers had found seven flaws that could allow indirect prompt injection — basically, malicious instructions hidden in a webpage or URL could trick ChatGPT into executing commands, bypassing safety controls, or even poisoning its memory.
NITDA at the time advised users to limit browsing and summarisation features on untrusted sites and only enable memory when needed.
Now apply that logic to your private messages. If ChatGPT can read and act on messages, what happens if someone sends you a message containing a hidden prompt injection? It's scenario security researchers are already debating.
For now, OpenAI says the Messages tool is read-first, with confirmed sends. But as AI assistants become more autonomous, the line between helpful and invasive is getting very thin indeed.
One to watch — and maybe one to think twice about before hitting "Allow".
Credit: cbinews.tv
#ChatGPT #Apple #iMessage #OpenAI #ApplePrivacy #MacOS #AI #TechNews #PrivacyMatters #Siri #AppleIntelligence #NITDA #CyberSecurity #CBINewsTV

