A Flaw in ChatGPT’s Mac App Could Have Let Hackers Grab Sensitive Data

Objective-See researcher Patrick Wardle reported a trivial but serious vulnerability in the ChatGPT macOS application: a trusted script interpreter can be abused to deliver untrusted scripts into the main ChatGPT process, enabling access to chat logs and the ability to execute commands in other applications. Wardle demonstrated a short proof-of-concept, has submitted the finding to OpenAI (which is reviewing it), and will present additional macOS AI-app bugs at an upcoming conference.