OpenAI Patches ChatGPT Mac App Vulnerability

OpenAI has patched a security vulnerability in its ChatGPT macOS application that could have allowed attackers to take control of the software on a victim's computer. As Wired first reported, the flaw would have granted an attacker access to stored chat logs, data, and interconnections such as browser sessions.

Researchers at the Objective-See Foundation discovered the bug, which bypassed digital signature checks intended to ensure only authorized OpenAI components could communicate. Patrick Wardle, a software analyst and macOS researcher at the Objective-See Foundation, explained that a trusted script interpreter could be manipulated to deliver untrusted scripts into the main ChatGPT process. Wardle described the exploit as "insanely trivial," noting that a proof of concept required only about a dozen lines of code.

While the vulnerability required an attacker to already have malware installed on the target machine, it could have enabled the software to run malicious commands that appeared as legitimate instructions.

On September 25, OpenAI documented the fix within its system change log. In a statement to Wired, OpenAI spokesperson Shane Bauer said the company continues to evolve its security practices but recognizes the "need to move faster."

Wardle, who will present further analysis on AI macOS bugs at the Objective by the Sea conference in November, cautioned that AI companies are currently fixated on adding features, which broadens the attack surface and often leaves security as an "afterthought."