AdvisoriesToolsStatsChecklistSubmitGet the email
← All advisories
VV-020highPrompt injection2024-09-20

SpAIware: prompt injection writes into ChatGPT’s long-term memory

Embrace The Red showed untrusted content planting persistent instructions in ChatGPT’s memory on the macOS app, so later chats kept sending data to the attacker.

01.

What it is

Published 20 September 2024. The attack chain uses prompt injection from untrusted data to insert long-term memories that keep exfiltrating conversations.

02.

How to check

  1. In ChatGPT, open Settings → Personalization → Memory and review stored memories.
03.

How to fix

  1. Delete memories you didn’t create; update the app.
04.

Sources

More advisories for this tool
The weekly email

One email a week. Only what changed.

New advisories, CVEs and fixes for Lovable, Bolt, v0, Replit, Base44 and Cursor apps. No roundups of things you already read.