Misbehavior1 sourceFirst seen
ChatGPT's Data Leak Gets a Half-Hearted Patch: Some Secrets Still Escape
Ourtake
The fix blocks many image-based leaks, but a chatbot that fetches whatever pixels an attacker writes into its replies is still a bug with extra steps.
What happened
A security researcher reported that ChatGPT's image markdown rendering could be abused through prompt injection to send conversation data to third-party servers. OpenAI began rolling out a client-side URL validation check on the web app that blocks many such requests but not all, and the iOS app had not received the fix.