Korean ministry finds harmful replies in AI character chatbots open without age checks
A government check of AI character chatbots usable without adult verification found some characters producing sexual and violent descriptions and replies encouraging self-harm, suicide or crime. The Ministry of Gender Equality and Family asked operators to strengthen youth protections.
Testing chats outside adult verification
The Ministry of Gender Equality and Family announced the results on Oct. 15. The check was carried out by the Youth Media Environment Protection Center, which the Korea Youth Counseling and Welfare Institute runs on the ministry’s behalf.
The center examined replies actually generated by characters on some services in the standard environment, without adult verification. In AI character chatbots, users pick or create a character with a personality and backstory and continue a conversation through generative-AI replies.
What the inspectors found
Reviewers applied the criteria for media harmful to minors under Article 9 of the Youth Protection Act and Annex 2 of its enforcement decree. They found detailed sexual content and depictions of sexual acts, violence and cruelty, and replies encouraging self-harm, suicide, crime or illegal acts. Some characters produced several types.
The release does not name the services that were checked.
What operators were asked to do
The ministry shared the findings with the operators and requested voluntary improvements. Characters likely to produce inappropriate conversations for minors should have their search, recommendation and chat scope adjusted according to adult-verification status.
The ministry also asked that seriously harmful content, such as encouragement of self-harm or suicide and help in carrying out specific crimes, not be generated for users of any age. The step is a request, not a sanction.
Follow-up checks
The ministry plans to continue working-level talks with operators and keep checking their self-inspections and safeguards. Youth policy director Yoon Se-jin said active efforts and ongoing management by operators are essential.
Lee Chang-ho, head of the center, said generative AI creates new content during each conversation, so protections must account for the replies themselves.
Sources & context
2026-10-15T23:46:17.709Z
Ministry of Gender Equality and Family · AI chatbot check (Korea.kr) ↗Ministry release · full text (hwpx) ↗Wikimedia Commons · CC0 ↗File photo of a hand holding a smartphone (2015); not a screen from any inspected chatbot. CC0 1.0.