A Flaw in ChatGPT’s Mac App Could Have Let Hackers Grab Sensitive Data (www.wired.com)

A recently discovered vulnerability in the macOS version of OpenAI’s ChatGPT could have allowed attackers to access sensitive user data, including chat logs and browser sessions, by essentially taking control of the app. Researchers from the Objective-See Foundation revealed that the vulnerability stemmed from a trusted script interpreter that could be manipulated to execute untrusted commands. This flaw was alarmingly easy to exploit, requiring only a handful of lines of code, which raises significant concerns about the security of AI applications and their inherent access privileges. This incident highlights the urgent need for AI companies to prioritize security, especially as they rapidly expand their software capabilities. OpenAI has acknowledged the issue and issued a patch, but experts like Patrick Wardle emphasize that as more features are added, the potential attack surface grows, making vigilant security practices critical. The case serves as a reminder for the AI/ML community to strike a balance between innovation and protective measures, ensuring that robust security mechanisms accompany the proliferation of AI technologies.