OpenAI
Misconfigured OpenAI agents posted 53 users' photos online
🧒 Discovery
Some of OpenAI's badly set-up robots showed everyone on the internet photos that belonged to real people, without ever asking for permission.
📰 Essential
Poorly secured OpenAI agents publicly posted 53 images belonging to users, without their knowledge or consent.
🔬 Expert
The incident highlights a recurring risk in autonomous agent setups: overly broad permissions or sharing settings can expose personal data with no human ever approving the publication. OpenAI wasn't aware of the leak until it was discovered, raising questions about how much real control exists over agents running in production.