"Oh? It Wasn't Real?" The Double-Edged Sword of the AI Era 'Deepfake' [Im Ju-hyung's Tech Talk]

Widespread Use of AI-Based Video Synthesis Technology 'Deepfake'
Positive Prospects for Whistleblower Protection and Film Industry
Concerns Over Criminal Use Such as 'Deepfake Pornography'

Who is the real one? A video synthesized using artificial intelligence deepfake technology featuring actor Alec Baldwin (left) and former U.S. President Donald Trump. Photo by YouTube capture.

[Asia Economy Reporter Lim Juhyung] Video editing technology using artificial intelligence (AI), known as 'deepfake,' can now be easily found anywhere. It is used as memes on YouTube or to create fake actors appearing in movies. However, as highly sophisticated deepfake technology that is indistinguishable from reality begins to spread, concerns are growing that it could be misused for fake news, pornography, and other malicious purposes.

Deepfake is a synthesis technology that uses a type of AI algorithm called Generative Adversarial Network (GAN). It is mainly used by overlaying computer-generated graphics (CG) of other faces or images onto a person's face.

Such synthesis was possible before, but it required manual work on each video frame, making it time-consuming and resulting in poor quality. However, deepfake allows AI to automatically synthesize videos through algorithms, enabling rapid processing.

Also, as websites supporting deepfake software functions like 'DeepFaceLab' and 'FakeApp' increase, it has become easier for the general public to access deepfake technology.

As deepfake becomes widespread in our society, its social impact is gradually increasing, and both its positive and negative effects are becoming clearer. This is why it is said to reveal the bright and dark sides of the upcoming so-called 'AI era.'

Regarding its positive effects, deepfake is very useful in protecting whistleblowers whose personal information exposure could threaten their jobs or lives.

The documentary "Welcome to Chechnya," released last June, is cited as a case where deepfake technology was used to protect human rights activists. / Photo by Internet Community Capture

One representative positive case of protecting whistleblowers using deepfake is the documentary "Welcome to Chechnya," released in the United States in June last year. This documentary exposes the Russian government's policy of persecuting sexual minorities, and there were concerns that the faces of LGBTQ+ human rights activists working inside Russia at the time could be revealed, risking their identities being exposed.

The production team came up with the idea of overlaying CG created by deepfake on the faces of the activists appearing in the footage, allowing the documentary to be safely and successfully filmed.

High-resolution deepfake technology released by US entertainment company Disney / Photo by Disney Research

Deepfake also shows great potential in the entertainment industry. Last year, the Disney Research Lab in the United States released an ultra-high-definition deepfake video with a resolution of one million pixels. This overcame the limitations of existing deepfake software, which showed obvious synthesis flaws on large screens due to low resolution, realizing the possibility of applying deepfake to commercial movie screens.

Using deepfake, it is possible to recreate a specific actor's younger or older appearance or even bring back deceased actors.

On the other hand, crimes using deepfake are increasing exponentially. Since last year, so-called 'deepfake pornography,' where celebrities' or ordinary people's faces are synthesized into pornographic videos using deepfake, has begun to be produced.

According to a report released on June 3 last year by Dutch cybersecurity research company Deeptrace, the number of deepfake videos online surged nearly twofold from 7,964 in 2018 to 14,678 in 2019. Among these, 96% (14,090) were deepfake pornographic videos.

Kang Hoon, known as 'Butta,' who was arrested for assisting Jo Joo-bin in producing and distributing sexual exploitation materials on the Telegram messenger, is being transferred from Jongno Police Station to the prosecution on April 17 last year. / Photo by Yonhap News

Deepfake pornography has already become a social issue in South Korea as well.

During the 'Nth Room' case, where individuals gathered on the messenger app Telegram produced and distributed sexually exploitative videos for money, some suspects committed so-called 'acquaintance defamation' crimes by synthesizing women's photos into other pornographic videos using deepfake, causing public outrage. Regarding Kang Hoon (18), who participated in producing and distributing such deepfake pornography, the prosecution requested a 30-year prison sentence on the 8th of last month.

Consequently, laws aimed at preventing the negative effects of deepfake are also being introduced. The amendment to the 'Special Act on the Punishment of Sexual Crimes,' which has been in effect since June last year, established provisions to punish crimes related to deepfake.

According to this law, producing or distributing pornographic material synthesized with faces or bodies using deepfake can result in imprisonment of up to five years or a fine of up to 50 million won. In particular, offenders who produce or distribute deepfake pornography for profit can face aggravated punishment of up to seven years in prison.

This content was produced with the assistance of AI translation services.