Deepfake dangers: how can Hong Kong schools protect students from image abuse?

Picture a birthday party. Children gather around a tablet, feeding a group photo into an app that makes everyone dance in silly hats and sing along. Or a family gathering, where a black-and-white picture of grandparents is uploaded into the same app, “reviving” them as young adults. It looks fun, harmless.

But according to Home Security Heroes, one in three of these apps can also generate pornographic content. The US-based identity theft protection company found in its 2023 State of Deepfakes report that 98 per cent of deepfake videos circulating online were pornographic, with 99 per cent of the victims women and girls. It takes less than 25 minutes and costs nothing, the report continued, to produce a 60-second fake video using just one clear face image. This speed and accessibility have outpaced not only parents’ awareness but also the capacity of governments and legal systems to respond and educate.

Maari Sugawara, an interdisciplinary artist, researcher and PhD candidate at City University of Hong Kong whose work focuses on critical AI studies, points to an institutional blind spot that affects places where technological innovation thrives, including Hong Kong.

“Legal problems begin when AI advancement is prioritised over the harm – usually misogynistic – that it can cause,” she says. In other words, innovation is institutionalised and backed by legal frameworks, but misuse is not met with the same level of accountability, enabling it to continue, often unpunished.

That reality hit the local education system last summer when a law student at the University of Hong Kong allegedly used free AI software to generate some 700 non-consensual pornographic images of 20 to 30 women, including the student’s university and primary school classmates and secondary schoolteachers, all based on social media photos. When the victims lodged a sexual harassment complaint with the Equal Opportunities Commission, they were told the case did not meet the legal definition, since the images were not intended for distribution – and so it could not proceed.

The government is looking to update existing laws before the end of the current administration’s term in June 2027. In the interim, a vacuum exists, and the burden of prevention, education and redress falls disproportionately on victims, NGOs and schools.

“We know it’s a problem concerning children and teenagers, whether they are targeted by classmates or adults,” says Jaap Marsman, adviser for safeguarding and child protection at the English Schools Foundation (ESF). “Hong Kong doesn’t have solid data about this crisis yet, but in 2024, the police released really worrying numbers showing that students as young as 11 accounted for a fifth of victims of ‘naked-chat blackmail’. More recent global reports show that we should absolutely be active about it.”