A viral social media trend encourages users to prompt ChatGPT to fill an empty room image with data it has collected about them. The chatbot populates the space based on stored memories, including names, preferences, and user goals. For Plus and Pro subscribers, the model also references past chat history to generate these results.
Security experts warn that sharing these results publicly on platforms like X exposes sensitive user profiles to strangers. This risk is particularly relevant for cryptocurrency holders, as stored information regarding specific holdings, exchanges, or travel plans can be leveraged by attackers to craft sophisticated phishing campaigns.
OpenAI provides options to view, delete, or disable memory features within the personalization settings, and offers a temporary chat mode that ignores stored data. Industry figures like Vitalik Buterin have proposed privacy-focused solutions, such as masking prompts, to prevent AI models from accessing sensitive information.
Regulatory scrutiny regarding data handling is increasing globally. A judge in New Mexico recently imposed a penalty of $25 billion on Meta due to its data management practices. As AI memory capabilities expand, users are increasingly using such visual prompts as a diagnostic tool to audit the information held by AI firms.
