Publicly available court records or news reports

ChatGPT macOS Flaw 'SpAIware' Allowed Data Theft and False Info Planting

Sep 25, 2024, 09:34 AM

Security researcher Johann Rehberger discovered a flaw in ChatGPT's memory feature that allowed attackers to plant false information and steal user data via indirect prompt injection. Rehberger reported the issue to OpenAI, which labeled it a safety issue rather than a security concern. Rehberger then created a proof-of-concept to demonstrate the flaw. The vulnerability, dubbed 'SpAIware,' could have enabled long-term spyware on the macOS version of ChatGPT, exposing user data across multiple conversations. The flaw has since been fixed.