AI researchers delete over 2,000 web links to suspected child abuse

Artificial intelligence (AI) researchers announced Friday (August 30) that they had deleted over 2,000 web links to suspected child abuse images from a dataset used to train common AI image generation tools.

An Associated Press report early Saturday said the dataset in the image was the Large-scale Artificial Intelligence Open Network (LAION).

LAION is a massive index of online images and captions that served as a source for leading AI image producers such as Stable Diffusion and Midjourney.

In December 2023, a report from the Stanford Internet Observatory said that LAION contained links to sexually explicit images of children, which contributed to the ease with which some AI tools could create photorealistic deepfakes depicting children.

This report prompted LAION to immediately remove its dataset.

Eight months after the report was published, LAION said it was working with the Stanford Internet Observatory and anti-abuse organizations in Canada and the UK to fix the problem and release a cleaned dataset for future AI research.

The Associated Press report said Stanford praised LAION for its significant improvements but said the next step would be to stop distributing the “contaminated models” that could still produce images of child abuse.

The cleaned-up version of the LAION dataset comes at a time when governments around the world are taking a closer look at how certain technological tools are being used to create or distribute illegal images of children.

(With contributions from agencies)