OpenAI says that it's encrypting chats on Mac's ChatGPT desktop app, after controversy

You can only download the app via OpenAI's website

Reading time icon 2 min. read


Readers help support MSpoweruser. We may get a commission if you buy through our links. Tooltip Icon

Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more

Key notes

  • OpenAI released its ChatGPT desktop app for Mac recently, with plans to expand to Windows later.
  • The app had a security flaw allowing plain text storage of conversations, as shared by an engineer.
  • Despite updating to encrypt chats, the app’s lack of sandboxing and potential privacy risks remain present.
ChatGPT desktop app

OpenAI opened up its ChatGPT desktop app for everyone on Mac not too long ago. Launched during the Spring Update event a little while ago, the Microsoft-backed AI company said that the hotly-anticipated desktop app will come to Windows after—which is an interesting choice of business.

But still, the ChatGPT desktop app (or, at least the Mac version for now) is plagued with a security flaw that allows plain text storage of conversations until recently. Engineer Pedro José Pereira Vieito shares on Threads, the popular “Twitter killer” app, how another application could easily access and display ChatGPT conversations right after they occurred.

That’s actually pretty concerning because you can imagine how vulnerable it makes your private conversations if any app can access and display them without encryption. Despite macOS security features since Mojave, which require explicit user consent for app access to private data, ChatGPT opted out of these safeguards (sandboxing).

The company managed to somehow find a leeway out of that requirement by only letting you download the app via the website, and not Apple’s App Store for Mac.

Following the controversy, OpenAI then responded by updating the ChatGPT app to encrypt local chats. But still, that doesn’t necessarily address the app’s lack of sandboxing which could still potentially bite back in the future.

Sure, OpenAI does use human/bot reviewers to train the model and you can opt out of it, but once the data falls into the wrong hands, chances are your conversations with ChatGPT could be exposed without your consent.