Overview

• Human Rights Watch identified over 170 images of Brazilian children in the LAION-5B dataset.
• These images were taken from personal blogs, social media, and video-sharing sites without consent.
• AI tools trained on these images have been used to create harmful deepfakes, including explicit content.
• LAION, the dataset's custodian, has pledged to remove the identified images but insists on shared responsibility for online content.
• Calls for stronger data protection laws and AI regulations are growing to safeguard children's privacy.