
On the 19th, I signed up for 'Crack' and 'Zeta,' AI character chat apps reportedly popular among teenagers. As soon as the apps opened, titles and images typically found on adult sites filled the home screen. Characters dressed in revealing clothing were also easily noticeable.
I entered my birth year as 2011, making me a minor aged 14–15 as of that day. Although both apps state they are only for users aged 14 and older, there is no procedure to verify the accuracy of the entered birth date, so users could sign up by simply entering any age. The barriers on app markets are even lower: Crack and Zeta are both distributed as 12+ on Google Play and the Apple App Store, meaning they can be downloaded by upper elementary school students.
Zeta distinguishes between 'Safe Mode' for general users and 'Unlimited Mode,' which requires adult verification to access. In Unlimited Mode, settings even include threats and sexual assault scenarios. Safe Mode is not limited to a 12-year-old level either. When I mentioned specific physical contact with Zeta's popular character 'Jung Ye-won' in Safe Mode, the AI halted the conversation. However, by avoiding certain keywords when inputting prompts, explicit text still appeared despite being in Safe Mode.
Crack presented even more serious issues. Although it implements 'Safety Mode' and 'Unsafe Mode' based on adult verification, deep physical contact was generated as text even within Safety Mode. This explains why the Korea Communications Commission has received a flood of complaints regarding explicitness and stimulation from AI chat apps.
Age ratings are determined not by national institutional investors but by app markets. While games receive pre-release ratings from the Game Rating and Administration Committee or self-classification operators, non-game apps are classified by Google and Apple based on information submitted by developers regarding explicitness, violence, etc. There is no procedure for the government to directly verify the level of conversation AI can generate before release.
AI conversations lack established review criteria like those for movies or webtoons. The same character can easily shift from daily conversation to sexual or violent scenes depending on user input and dialogue flow. While output levels vary by model or safety mechanisms, no separate review criteria for AI chat apps are confirmed in publicly available app market policies.
Users who judge the age rating to be lower than the actual service must report it to Google or Apple. Illegal information or content harmful to minors can also be separately reported to the Korea Communications Commission. However, all measures are merely reactive responses after problems occur, offering no preventive Jo Chi-neun (EVP). The gap between the 12th+ rating label and conversations generatable from actual minor accounts is large, making urgent preparation of alternatives necessary.
