OpenAI’s Mac app stored conversations in plain text

OpenAI’s Mac app contained a potential security flaw, which may have allowed malicious apps to access users' conversations with ChatGPT.

OpenAI's Mac app, made available for users last week, was storing information in plain text. This means that if a hacker or malicious app gained access to a Mac device, it would have the ability to read users' conversations with ChatGPT without any permission prompt.

Pereiro Vieta uncovered this issue on Threads, where he demonstrated that an app can access conversations with ChatGPT.

After the problem was raised, OpenAI issued an update.

"We are aware of this issue and have shipped a new version of the application which encrypts these conversations," OpenAI spokesperson Taya Christianson said in a statement to The Verge. MacOS users can now download the latest version of the chatbot to avoid potential vulnerabilities.

As Vieto pointed out in his post, macOS has blocked unauthorized access to any user's private data since macOS Mojave 10.14, which was released six years ago.

This means that any app, such as Calendar, Photos, or Mail, can access private user data only with explicit user permission.

However, OpenAI chose to opt out of Apple's security control system and store the conversations in plain text in a non-protected location, disabling all of these built-in defenses.

Last month, Apple announced a partnership with OpenAI, allowing users to access ChatGPT software on Apple's devices with their permission.

While the agreement is separate from ChatGPT releasing an app for Macs, it may also cause further concerns about how securely OpenAI will handle users' data. Some cybersecurity experts say that ChatGPT integration in Apple’s software potentially adds additional attack vectors.