HRW Finds Australian Children's Photos Misused in AI Training, Raises Privacy Concerns
TEHRAN (Tasnim) – Human Rights Watch (HRW) has uncovered alarming misuse of photos of Australian children in AI training datasets, highlighting serious privacy risks despite families' strict online privacy settings.
Last month, HRW researcher Hye Jung Han identified 190 photos of children from all Australian states and territories in the popular AI dataset LAION-5B. These photos were included without the knowledge or consent of the children or their families, potentially exposing them to privacy violations, according to ArsTechnica.
The dataset spans the entirety of childhood, enabling AI image generators to create realistic deepfakes of these children. Han's report noted that some URLs in the dataset reveal identifying information, including names and locations, which could facilitate tracking of children otherwise hidden online.
Han highlighted specific cases, such as a photo showing two boys in Perth, where she traced their full names, ages, and even the name of their preschool from a single image link. This discovery underscores the inadequacy of current privacy protections, as even photos posted with stringent settings on platforms like YouTube were not safe from being archived and used.
Despite policies prohibiting AI scraping, platforms like YouTube have struggled to prevent unauthorized use of children's data. YouTube's spokesperson, Jack Malon, emphasized the platform's efforts to combat such abuses but acknowledged the challenges in completely removing archived content from datasets.
Han's findings come amidst impending reforms to Australia's Privacy Act, including the introduction of the Children’s Online Privacy Code. However, uncertainties remain about the scope of these reforms and their effectiveness in protecting children's digital privacy.
The AI community faces criticism for its handling of personal data, with calls for stricter regulations to prevent nonconsensual use of children's images in AI systems. HRW has urged policymakers to implement robust safeguards that prohibit scraping of children's data and provide effective remedies for those affected by such misuse.
As AI technology continues to evolve, Han emphasized the urgency of protecting children's data privacy to prevent further exploitation of their images. She stressed that proactive measures now could shape a future where AI technology respects and upholds children's rights online.