The AI database LAION-5B, used for training AI image generators, has been taken down due to the discovery of over 3,000 child sexual abuse images.

The Stanford Internet Observatory and anti-abuse organizations identified and reported the illegal images.

AI models trained on LAION-5B have been used to create and distribute realistic child abuse images on the dark web.

LAION, the nonprofit maintaining the dataset, has temporarily removed it for safety checks before republishing.

There are concerns about the existence of other AI models trained on LAION-5B that generate explicit imagery.

The Stanford University report recommends the removal of all tools created using LAION-5B and accountability for platforms hosting older versions.

The report also calls for measures to clean up the data and ensure older versions of AI tools disappear from the internet.