June 17, 2024
Cited article by Human Rights Watch
HRRC condemns the nonconsensual use of Brazilian children’s personal photographs and information for the development of artificial intelligence (AI) tools and urges the Brazilian government to enforce stronger data privacy protections for children.
News Brief
In a recent report, Human Rights Watch found that LAION-5B, a large data set managed by German nonprofit LAION and used to train popular AI models, contains personal photos and details of Brazilian children. Photos originally posted by children and their families on personal websites and media are being scraped from the internet into LAION-5B—and consequently, widely used AI tools and programs—without the knowledge or consent of the children.
This critical breach of privacy carries serious implications for children’s safety, for it poses tangible risks of digital sexual exploitation. The report cites that in the past, people have utilized AI models trained with LAION’s data to create realistic explicit images based on photos of real children. Alarmingly, photos of child sexual abuse obtained from the web and stored in LAION-5B have also been used to produce explicit images of abuse victims.
The report criticizes Brazil’s General Personal Data Protection Law for not ensuring effective protections for children and their data privacy. In addition, the report calls for Brazil’s National Council for the Rights of Children and Adolescents and the Ministry of Human Rights and Citizenship to enact a new national policy that bans the nonconsensual collection, replication, and manipulation of children’s personal data and likenesses by AI systems.