AI-generated child sexual abuse images could flood the internet. Now there are calls for action

36

The escalating problem of child sexual abuse images on the internet could be exacerbated by unregulated artificial intelligence tools capable of producing deepfake photos, according to a warning from a monitoring group on Tuesday. The Internet Watch Foundation, based in the U.K., has released a report urging immediate action from governments and tech companies to prevent an inundation of AI-generated child sexual abuse images, which could overwhelm law enforcement and increase the number of potential victims.

Dan Sexton, the group’s chief technology officer, emphasized the urgency of the situation, stating that the issue is current and demands immediate attention. In a groundbreaking case in South Korea, an individual was recently sentenced for using AI to create numerous virtual child abuse images.

The report also highlights instances where children themselves have used these tools harmfully. In southwestern Spain, police are investigating a case where teens allegedly used an app to create nude images of their classmates.

This alarming trend underscores the negative aspects of the race to develop generative AI systems, which are designed to create outputs ranging from text to artwork based on user inputs. The unchecked proliferation of deepfake child sexual abuse images poses significant challenges for investigators and could be used to groom new victims.

The report reveals that famous children’s faces have been found online, along with a high demand for creating more images of children who have previously been victims of abuse. Sexton noted the shocking use of existing content to generate new abusive images.

The Internet Watch Foundation began receiving reports about AI-generated abusive imagery earlier this year, leading to investigations into dark web forums. Abusers were found sharing tips and exploiting their home computers to create lifelike sexually explicit images of children.

While the IWF’s report aims to raise awareness rather than provide solutions, it calls for governments, especially in the European Union, to strengthen laws against AI-generated abuse. The report also highlights the need for technology providers to prevent their tools from being misused, acknowledging the complexity of reining in already released technologies.

Most newly introduced AI image-generators, including OpenAI’s DALL-E, contain mechanisms to block the creation of child sex abuse material. However, the open-source tool Stable Diffusion, developed by Stability AI, has been widely used for creating explicit content, including nonconsensual imagery. Stability AI has since implemented new filters and licensing terms to combat illegal uses.

David Thiel of the Stanford Internet Observatory notes that older versions of Stable Diffusion remain popular for creating explicit child content. The IWF report recognizes the challenges in criminalizing AI image-generating tools and stresses the difficulty of regulating private computer use.

Existing laws in the U.S., U.K., and elsewhere typically deem AI-generated child sexual abuse images illegal, but it’s uncertain if law enforcement is equipped to tackle this issue. Ian Critchley of the National Police Chiefs’ Council highlights the disturbing trends observed by officers in identifying victims.

The IWF’s report is being released ahead of a global AI safety summit, aiming to discuss the darker aspects of AI technology with key figures, including U.S. Vice President Kamala Harris and tech leaders. IWF CEO Susie Hargreaves remains optimistic despite the grim situation, emphasizing the importance of awareness and dialogue.

This report includes contributions from O’Brien in Providence, Rhode Island, and Big Big News writers Barbara Ortutay in Oakland, California, and Hyung-jin Kim in Seoul, South Korea.