Updated 2 November 2026: checked against Korean reports that quote the bill, now linked, because we could not open the bill on the National Assembly's site. We corrected the account of the 2025 Goyang acquittal: reports say prosecutors failed to prove the woman existed, not that the law could not apply to synthetic images. We added the proposed wording, which covers a fictional person "who can be recognised as a real person", and the 18,000 public comments. We removed "one of the world's toughest regimes", "a rare choice among democracies", the Japan comparison and the compliance checklist, none of which we could support. We added that the Constitutional Court upheld the obscenity ban in the decision we cite.

Correction (19 October 2026): An earlier version said the bill was referred to the subcommittee in August 2026 (it was put on the Legislation and Judiciary Committee's agenda and referred on 27 July 2026) and attributed the crime statistics to the bill's sponsors (they come from a July 2026 Korean Women's Development Institute report); the descriptions of California's SB 1276 and of the EU directive have also been made more precise.

A bill led by Democratic Party lawmaker Huh Young with nine co-sponsors would add a new Article 14-4 to the Act on Special Cases Concerning the Punishment of Sexual Crimes. It was filed on 11 September 2025 and, on 27 July 2026, was put on the Legislation and Judiciary Committee's agenda and sent to its subcommittee. It is not law, and it may be amended or dropped. We found no report of a subcommittee vote as of 2 November 2026.

What does the Korean bill say?

The bill would punish AI-made sexual images of "a fictional person or representation who can be recognised as a real person". We could not open the bill text, so this description relies on Weekly Kyunghyang, which quotes the article, and on The Korea Times and Hankook Ilbo.

  • Creating such material in a form that can arouse sexual desire or shame: up to seven years in prison or a fine of up to 50 million won.
  • Distributing, selling, renting, providing, displaying, screening or advertising it: the same penalty.
  • Possessing, buying, storing or viewing it: up to three years or a fine of up to 30 million won.

These figures match the existing deepfake offence in Article 14-2. According to Weekly Kyunghyang, the bill does not require a resemblance to a specific person, a real person harmed, or an act against anyone's will, and private creation could be caught. During the legislative notice period of 17 to 26 September 2025, Hankook Ilbo reports about 18,000 comments on the National Assembly's system, mostly criticising it as excessive control of free expression.

Why was it proposed?

The sponsors cite misuse of AI and a legal gap. The Korea Times reports figures from a July 2026 study by the state-funded Korean Women's Development Institute: reported digital sex crimes rose from 9,385 in 2015 to 15,612 in 2024, and deepfake sex crimes from 31 in 2020 to 550 in 2024. Those figures cover all digital sex crimes and deepfakes, not fictional-person material.

Rep. Huh also points to a 2025 acquittal by the Goyang branch of the Uijeongbu District Court. A man had forwarded AI-composited nude images in a Telegram chat. Reports differ on the month (July in Hankook Ilbo, August in Weekly Kyunghyang). Hankook Ilbo quotes the court: it was hard to conclude that the unidentified person in the charge exists in reality. Weekly Kyunghyang adds that prosecutors did not appeal and that the defence itself warned against reading the case as meaning AI images of unreal people are always acquitted. The defence also said a different obscenity charge might have changed the result. So the case shows a problem of proof, and whether it requires the broader offence is part of the debate.

What does Korean law already cover?

  • Real people. Article 14-2 covers images of a person's face, body or voice edited against that person's will. A 2024 amendment added possession, buying, storing and viewing. Our analysis of Korean deepfake law has the background.
  • Children. The child protection act covers a person or representation clearly recognisable as a child or youth, as the Supreme Court explained in 2025.
  • Obscenity. According to Kyunghyang, distributing obscene material of virtual persons can already be punished under the Information and Communications Network Act, by up to one year in prison or a fine of up to 10 million won.
  • Labelling. The AI Basic Act, in force from 22 January 2026, is a transparency law. It does not make such material an offence.

Courts also continue to apply Article 14-2 to new facts; see our note on a Seoul conviction over clothed images.

Can there be a crime without a victim?

That is the central objection, and Korean commentators are divided.

In favour. Rep. Huh told Weekly Kyunghyang that the aim is to fill a gap and prevent harm as such material is sold online. A representative of the Korea Sexual Violence Relief Center told Kyunghyang that claiming a subject is unreal is a common evasion, and that such images are mostly built by recombining real women's bodies and faces.

Against. A law professor told Hankook Ilbo that with no victim it is unclear whose shame the test refers to. A National Assembly Research Service researcher told the same paper that existing law can already punish distribution and that applying the sexual-crimes statute is excessive. Lawyers quoted by Kyunghyang raised freedom of expression and suggested using the network act instead. Weekly Kyunghyang also reports criticism that the wording "can be recognised as a real person" is unclear, says it could cover a wholly fictional photorealistic figure, and asks whether penalties copied from the deepfake offence suit making, for-profit distribution and mere viewing alike.

A constitutional challenge is possible if the bill passes, in our view. In a 2009 decision (2006Hun-Ba109) the Constitutional Court said obscene expression falls within protected speech. It still upheld the criminal ban on distributing it online. So protection does not mean the bill would fail. The question would be proportionality, particularly for mere viewing.

How do other legal systems treat fictional sexual material?

Comparable rules exist, but they are usually narrower. These are the examples we checked; this is not a survey.

In our reading, the Korean bill stands out in the examples above because it reaches adult-appearing material with no listed category of extreme acts, and reaches viewing as well as possession.

How would it be proved?

In our view the bill changes the evidence problem but does not remove it. Prosecutors would have to show that an image is AI-made and could be taken for a real person. Provenance tools can help, but metadata can be stripped, as we discuss in our note on marking AI-generated content. The Goyang case itself turned on missing evidence of the image's source, according to Weekly Kyunghyang.

What does it mean for Türkiye?

We know of no Turkish provision written for AI-made sexual images of fictional adults. The Turkish Penal Code already has obscenity offences in Article 226. These include publishing obscene images in the media (paragraph 2) and producing them using children, representative images of children or people who look like children (paragraph 3). In our reading, those provisions do not depend on a real person being harmed. Access to such content can be blocked under Article 8 of Law No. 5651.

Images of real people are a separate matter: Articles 134 to 136 of the Penal Code, personality rights under the Civil Code (Articles 24 and 25) and access blocking for private life under Article 9/A of Law No. 5651. This site has published its own 100-article proposal for a Turkish AI act, which has no official status.

What to watch next

Weekly Kyunghyang expects the subcommittee to focus on whether identifiability and real harm become elements, and whether making, distributing and viewing get separate penalties. Hankook Ilbo notes the bill may be amended or dropped. Rep. Huh's office told it the free-expression arguments will be discussed in review.

Frequently asked questions

Has the Korean AI sexual content bill become law?

No. It was filed on 11 September 2025 and sent to the Legislation and Judiciary Committee's bill-review subcommittee on 27 July 2026. We found no report of a vote as of 2 November 2026, and the text may change.

Would the bill punish merely viewing such material?

As reported, yes. Possessing, buying, storing or viewing would carry up to three years in prison or a fine of up to 30 million won, matching Article 14-2(4) of the Korean sexual-crimes act for deepfakes of real people. Weekly Kyunghyang calls this an especially contested part of the bill.

Would anime or drawings be covered?

That is unclear. The reported wording covers a fictional person or representation "who can be recognised as a real person", made using AI technology. That suggests realistic images, but the text does not define the test, and Korean commentators have called it vague. The final wording will decide.

Aren't fictional depictions of children already illegal in Korea?

Korea's child protection act covers a person or representation clearly recognisable as a child or youth. The bill's new reach is adult-appearing fictional persons. Kyunghyang reports that some cases fall outside the child act because the subject appeared to be an adult.

Is there a comparable rule in Türkiye?

We know of none written for AI-made sexual images of fictional adults. Article 226 of the Turkish Penal Code covers certain obscene material, and images of real people fall under private-life, personal-data and personality-rights rules.

Burhan Doğuş Ayparlar's View

This section reflects my personal assessment as the founder of this site and an attorney at law (Türkiye).

In my view, the bill identifies a real problem but, in part of its remedy, strains the limits of criminal law. Generative models can weaken a system that depends on proving a real person. The Goyang case shows the difficulty, although it was a case of missing proof and not one of a settled legal rule.

Punishing creation and distribution is not the same as punishing the possession or viewing of adult-appearing content with no concrete victim. The latter carries serious risks for legal certainty and proportionality, and I think it could move criminal law from harm towards moral policing. A more balanced model would target tools and distribution chains that facilitate non-consensual sexualisation.

The lesson I draw for Türkiye is that rules on deepfakes should be built around where concrete harm arises. In my view, fast and effective protection against non-consensual synthetic images of real people should come first.

Sources: The Korea Times (11 August 2026), Weekly Kyunghyang (29 July 2026), Hankook Ilbo (6 August 2026), Kyunghyang (3 August 2026), Ministry of Justice notice on the 2024 amendment; Korean Constitutional Court, 2006Hun-Ba109 (2009), English summary; US, UK and EU texts as linked; Turkish laws on mevzuat.gov.tr. Read on 1 and 2 November 2026. We could not open the bill or its number on the National Assembly's system, so its wording comes from those reports, translated by us. The bill may change. General information, not legal advice.