A recent report by Human Rights Watch (HRW), an international human rights organization, reveals that popular AI image generators have been trained on datasets containing numerous personal photos of Brazilian children without permission. This poses a serious challenge to children's privacy and safety.
The report indicates that HRW researchers discovered at least 170 personal photos of Brazilian children from 10 states within the massive public dataset "LAION-5B," covering precious moments from their entire childhood, from infancy to adolescence. Most of these photos were originally shared among friends and family on personal blogs and videos, with no intention of public exposure.
Image Source Note: The image is generated by AI, provided by the image licensing service Midjourney
Notably, many of these photos contain children's names and location information, making them highly traceable and identifiable. Misuse of these images could expose children to risks such as bullying and phishing. Worse still, some analyses suggest that with the advancement of AI face-swapping technology, children's portraits could be used to create disturbing explicit content.
The report cites instances where at least 85 Brazilian girls have reported that their classmates used AI tools to create and disseminate fake pornographic videos of them online, causing lasting psychological trauma. HRW calls on the government to take immediate action to protect children's data from being misused by AI systems.
In response, the LAION organization has collaborated with HRW and other institutions to remove URLs linking to child images from the training data. However, the report warns that this may "significantly underestimate the actual total amount of children's personal data in the dataset."
HRW researchers note that in the past, these family photos were somewhat "private" on the internet, less likely to be discovered. But now, with their extensive use by artificial intelligence, they have breached the original boundaries of privacy. She criticizes the unfairness of shifting the responsibility of protecting children's privacy to parents, arguing that governments and tech companies should take responsibility and implement protective measures.
Human Rights Watch calls on the Brazilian government to explicitly prohibit the unauthorized use of children's personal data for training AI systems in the upcoming child rights protection regulations, and to establish avenues for accountability for affected children. Some also advocate for stronger legal protection of children's data globally to prevent misuse in AI systems or the creation of inappropriate content.