What Is a Deepfake?
A deepfake is a human image synthesis technique that uses deep learning, a branch of machine learning.
Put simply, it is a technology that combines photos of real people to create edited video content.
As deepfake technology has advanced, even ordinary people who are not experts can easily use it to create content, and this has led to problems such as combining it with nude photos to produce and distribute obscene material, or using it for scams like phishing.
This post focuses on crimes such as creating and circulating fake obscene material.
Deepfake Crimes
Deepfake crimes have long targeted celebrities and others. Recently, however, the issue drew even more attention after Telegram deepfake sex crimes revealed large-scale offenses not only at more than 100 universities but also at middle and high schools, and within the military.
Deepfake Punishment Provisions
The provisions for punishing deepfakes are set out in Article 14-2 of the Act on Special Cases Concerning the Punishment of Sexual Crimes (성폭력처벌법)
Article 14-2 (Distribution, etc. of False Video Material, etc.) (1) Any person who, for the purpose of distribution, etc., edits, synthesizes, or processes (hereinafter referred to as "editing, etc." in this Article) a filmed item, video item, or audio item targeting another person's face, body, or voice (hereinafter referred to as "video item, etc." in this Article) into a form that may cause sexual desire or shame against the will of the subject of the video item, etc., shall be punished by imprisonment for not more than 5 years or by a criminal fine not exceeding KRW 50 million.
(2) Any person who distributes, etc. an edited item, synthesized item, or processed item under paragraph (1) (hereinafter referred to as "edited item, etc." in this paragraph) or a reproduction thereof (including a reproduction of a reproduction. The same shall apply hereinafter in this paragraph), or who, even where it was not against the will of the subject of the video item, etc. at the time of the editing, etc. under paragraph (1), subsequently distributes, etc. such edited item, etc. or reproduction against the will of the subject of the video item, etc., shall be punished by imprisonment for not more than 5 years or by a criminal fine not exceeding KRW 50 million.
(3) Any person who, for profit, commits the offense under paragraph (2) through an information and communications network against the will of the subject of the video item, etc. shall be punished by imprisonment for not more than 7 years.
(4) Where a person habitually commits any of the offenses under paragraphs (1) through (3), the penalty prescribed for the relevant offense shall be aggravated by up to one-half. <Newly Inserted on 2020. 5. 19.>
[This Article Newly Inserted on 2020. 3. 24.]
Punishment for Viewing Deepfakes?
There is no provision punishing the mere viewing of deepfakes.
Under Article 14 of the Act on Special Cases Concerning the Punishment of Sexual Crimes, a person who views, purchases, or possesses material constituting the offense of filming using a camera, etc. may be punished as well; however, Article 14-2 of the Act on Special Cases Concerning the Punishment of Sexual Crimes does not provide for the punishment of persons who view, purchase, or possess such material.
That said, if the subject is a minor, a person may be punished for mere viewing under the Act on the Protection of Children and Youth Against Sex Offenses (청소년성보호법)(commonly known as the "Achungbeop").
Article 11 (Production, Distribution, etc. of Sexually Exploitative Material of Children and Youth) (5) Any person who purchases sexually exploitative material of children and youth, or who possesses or views such material knowing that it is such material, shall be punished by fixed-term imprisonment for at least 1 year.
Elements of a Deepfake Offense
The most important element of a deepfake offense is that there must be a purpose of distribution, etc.
In other words, the mere act of creating the material is difficult to punish, whereas distributing it can be punished.
However, if the material was not distributed and was created only for personal possession, proving the purpose of distribution is extremely difficult.
In practice, actual cases of punishment for deepfake obscene material are quite rare.
Deepfake Punishment
Punishment for deepfakes may be possible in the following cases.
- Requesting the creation of a deepfake
If the creator distributes the video, the requester may be classified as an accomplice and become subject to punishment. - Downloading via torrent
If you download using a torrent, the file is shared with others, so you may become a distributor and be punished. - Synthesizing and viewing deepfakes of minors
Deepfakes targeting minors can be punished under the Act on the Protection of Children and Youth Against Sex Offenses. - Sharing deepfake videos
If you share a deepfake video, you can be punished under Article 14-2 of the Act on Special Cases Concerning the Punishment of Sexual Crimes.
Security Dispositions for Deepfakes
If you receive a disposition of a criminal fine or heavier as a result of a deepfake, you may be subject to a security disposition in addition to the criminal punishment.
A security disposition is imposed when a person receives a disposition of a criminal fine or heavier for a sex crime.
Contents of Deepfake Security Dispositions
The contents of deepfake security dispositions are as follows
1. Disclosure of personal information
2. Completion of sex education
3. Employment restrictions
4. Visa restrictions
5. Wearing an electronic ankle monitor, etc.
Countermeasures Against Deepfake Crimes
Countermeasures against deepfake crimes are being prepared on multiple fronts.
DeepBrain AI's deepfake detection solution, Intel's FakeCatcher, and others have introduced technologies capable of determining the authenticity of deepfake content.
Naver's AI chatbot CLOVA X has applied safeguards to its image editing feature so that generated images cannot be used to create content such as obscene material, and companies such as Kakao Brain, TikTok, and Instagram will label AI-generated content as such.
The government is also preparing countermeasures.
Through its situation room, the Korea Communications Standards Commission receives reports of harm from victims and support organizations and conducts monitoring, holding daily electronic deliberations to order corrective measures for digital sex crime material within 24 hours.
The Ministry of Gender Equality and Family plans to build a system by 2025 that uses AI to find deepfake obscene material and automatically requests online operators to delete it.
A bill imposing an obligation to label AI-generated virtual content as AI-generated has been proposed and is currently pending in the National Assembly.
Attorney Specializing in Sex Crimes
As shown above, deepfake crimes are very serious sex crimes, and countermeasures are being prepared at the government level. Therefore, even if a deepfake crime is not punished immediately, it may be punished later, so you should refrain from committing such crimes, and if you become involved in a related offense, it is advisable to seek the help of a criminal defense attorney specializing in sex crimes.
