Reporting · technology
Viral ChatGPT memory prompt can reveal stored user data
A viral ChatGPT prompt asking the chatbot to fill an empty room can surface stored user data, raising privacy and phishing concerns for crypto users.
Quick take
- A viral ChatGPT prompt can bring up data the chatbot has stored about users.
- OpenAI says its memory feature holds saved memories and chat history for users.
- Security concerns include phishing, since stored holdings or exchanges could help scammers.
Crypto outlet BeInCrypto flagged a social media trend in which people try one prompt asking ChatGPT to fill an empty room, a request that can surface details the bot has kept about them. A popular crypto account on X urged its followers to try it. Images posted publicly on X could expose the profiles of the people who share them. The prompt has spread quickly across social feeds.
OpenAI says its memory feature holds saved memories and chat history, and that Plus and Pro subscribers also get a chat history reference. According to the company, users can view, delete, or disable memory.
An earlier case this year involved Claude chats containing wallet data that Google indexed. Stored details such as holdings or exchanges could help phishing scammers, and Changpeng Zhao has argued that the AI danger lies in phishing.
Vitalik Buterin has built a paranoid mode that masks AI prompts.
Sources
Machine-written from the publishers’ reports and checked against them · How Dextape writes
Mentioned in this story
CoinGecko last prints. Not a Dextape quote.
Reporting
Elliptic traces $124 million in stolen Binance assets converted to Bitcoin
November 8, 2026 · 1 min read
Reporting
Binance hot wallet theft estimated at $6.3 billion, analysts link it to North Korea
October 30, 2026 · 2 min read
Reporting