TiU researchers warn The Hague: deepfakes dangerous for democracy
In five years, as much as ninety percent of all online content will be partially manipulated or even a complete deepfake. Moreover, it will become increasingly difficult to distinguish real videos and photos from fakes. In a recent report, TiU researchers warn the government of an insidious future. “Don’t intervene only after the harm has already been done.”
Admittedly, the TikTok video in which the English royals supposedly sing along with O-Zone’s classic Dragostea Din Tei is very witty. But a world where manipulated videos are only used for satirical purposes is a utopia, warns TiU associate professor Bart van der Sloot. At the request of the Ministry of Justice and Security, he conducted research on the dangers of deepfakes together with Yvette Wagensveld and Bert-Jaap Koops, all affiliated with the Tilburg Institute for Law, Technology, and Society.
“In The Hague, people are concerned about the impact of manipulated digital material on democracy,” Van der Sloot explains about the reason for the research. “In the House of Representatives, the subject has already been discussed a few times as a result. Currently, detectors are still able to recognize 65 percent of manipulated images. According to experts, that percentage is decreasing rapidly. We have, therefore, mapped out what is happening in the field of deepfakes and how dangerous they are.”
Deepfakes are fake images, sounds, or photos, circulating online. They are edited or even completely fabricated and do not represent a true situation. However, they are often realistic and, due to rapid technological advances in artificial intelligence and machine learning, increasingly difficult to distinguish from the real thing. Based on existing videos and images, algorithms create new moving images of situations and even people. Add in sound clips, and Prime Minister Mark Rutte says in this way that “the coronavirus crisis was child’s play.”
The researchers see that problems already exist in several political and social areas in the Netherlands due to the use of deepfakes: from manipulated evidence finding its way into court cases to interference in elections. These are only expected to increase in the coming years. The researchers thus conclude that deepfakes have the potential to affect democratic processes.
And there is more. Van der Sloot: “Perhaps the biggest social problem is that deepfake is currently used primarily to make porn videos. Videos in which women appear to be performing sexual acts that they have not done at all. When teenagers do this to their classmates, it has a huge impact on young girls growing up and their social status. Also, for women in public office, on top of all the threats they face, this is a huge problem.”
Call for discussion
Dutch legislation in this area could be more proactive, says Van der Sloot. “Spreading misleading information is punishable, just like making porn without someone’s permission. Only, in that case, the video or image is already there, with all the disruptive consequences that this entails. The question is whether the existing regulation comes too late. In our report, we call for awareness, both among citizens and the government, of the dangers. But we also point out that the time has come to have a political discussion about the desirability of other legislation.”
Other countries such as China and the United States have advanced deepfake legislation. Particularly the latter is leading the way, according to Van der Sloot. “Certain states have a ban on making deepfakes during election time.” Whether that approach has political support here, he does not know. “Our government tends to appeal to the responsibility of the Dutch themselves. A ban on deepfake technology is drastic. You quickly end up in a legal quagmire, where discussions are held about whether that is censorship. And that is exactly the discussion that I encourage.”