Hacking ChatGPT's Memories with Prompt Injection

# Hacking ChatGPT's Memories with Prompt Injection

https://gcdradar.vercel.app/incidents/hacking-chatgpt-s-memories-with-prompt-injection · AIR-2025-0064 · closed 2025-11-21

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2025-11-07 (happened 2024-02)
- AI role: Attack on AI
- Products: ChatGPT

## Summary

A researcher demonstrated that ChatGPT's memory feature can be poisoned using prompt injections hidden inside shared Google Docs. The injected false facts successfully persisted across multiple chat sessions.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-11-07: ChatGPT: Hacking Memories with Prompt Injection (https://embracethered.com/blog/posts/2024/chatgpt-hacking-memories)
- 2025-11-07: Hacking ChatGPT's Memories with Prompt Injection (https://atlas.mitre.org/studies/AML.CS0040)

## Sources

- 2025-11-07 MITRE ATLAS: Hacking ChatGPT's Memories with Prompt Injection https://atlas.mitre.org/studies/AML.CS0040
- 2025-11-07 Embrace The Red: ChatGPT: Hacking Memories with Prompt Injection https://embracethered.com/blog/posts/2024/chatgpt-hacking-memories

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.