Many Danes doubt they can spot AI-made sexual images

A survey conducted by Epinion for DR found that 41 per cent of Danes say they would probably not be able to recognise manipulated sexual material, including so-called deepfake pornography. A further 32 per cent are unsure whether they could identify it.

Advances in artificial intelligence have made it harder to determine what is real and what is fake, several respondents said.

“I am sure I am being deceived every day, and the technology is only getting better,” said Peder Hammerskov, head of the Center for AI at the Danish School of Media and Journalism.

AI tools can place anyone in a sexual situation they have never experienced and never consented to. All that is needed is a photograph of a real person. Within seconds and a few clicks, it can be turned into a fake nude image or video.

Sexualised AI images pose particular risks

The technology used to create fake nude images and videos is developing rapidly. Ask Hesby Holm, director of the organisation Digitalt Ansvar, said it was therefore unsurprising that Danes doubted their ability to identify manipulated content.

“We need to be able to navigate facts in our public debate. That becomes difficult when we can no longer necessarily agree on what is true or false,” Holm said.

“Historically, we are at a very critical point. Previously, we have been able to distinguish between truth and falsehood. We can no longer do that because the technology has become so good.”

Holm said the consequences could be particularly serious when people could no longer recognise sexual material manipulated with AI.

“Generally, when it concerns sexually manipulated material, it is a particular problem,” he said.

“It is enormously invasive and humiliating to be placed in a sexual context. You lose the right to your face, your reputation and your own narrative. Once it has been shared, it is extremely difficult to have it removed again. That means it is completely outside your control.”

Experts no longer trust their own eyes

The 41 per cent of Danes who believe they would probably not be able to recognise fake nude images or videos created with artificial intelligence are not alone.

Hany Farid, who teaches at Dartmouth College in the United States and is one of the world’s leading researchers on deepfakes and AI, also no longer trusts his own eyes.

“No one can tell the difference anymore. In a year or two, our ability to trust what we see will be completely gone,” he said in an interview with The New York Times.

Hammerskov said he had the same experience.

“I am sure I am being deceived every day, and the technology is only getting better,” he said.

A growing phenomenon

In Denmark, it is illegal both to produce and share material showing real people manipulated into a sexual context without their consent.

Despite this, technology capable of producing such material is available in several places, including so-called nudify apps. These apps and websites use AI to create fake nude images of real people. They can remove clothing, make a picture move in a pornographic manner or place a person’s face on a naked body.

Last year, experts from Oxford University found at least 35,000 apps freely available on the internet that could create deepfakes, including videos manipulated with artificial intelligence.

“Many models are labelled with tags such as ‘porn’, ‘sexy’ or ‘nude’, or have descriptions indicating that they are intended to generate intimate images without consent,” the experts wrote in their report.

In 2023, the EU estimated that the number of deepfakes would rise from 500,000 in 2023 to eight million in 2025. It also estimated that 98 per cent of all deepfakes were pornographic.

“Although we lack comprehensive global figures, I am certain that the number of pornographic deepfakes has risen explosively,” Hammerskov said.