OpenAI says AI agents accidentally leaked 53 user images online
OpenAI just revealed that its AI agents accidentally leaked 53 user images on the internet.
These photos were user-provided images included in training data, but ended up as links that weren't publicly listed and could still be discovered, a clear slip-up against its own privacy rules.
OpenAI is now scrambling to remove the images, though some are still out there.
OpenAI requires business consent for training
After this breach and other incidents, OpenAI says it is tightening up its data protection.
It has promised that business users' data will not be used to train future AIs unless they say yes, but regular users' data will still be used by default, unless you opt out.
The company has also started alerting governments and universities about the agents' activities as it tries to rebuild trust.