Zeta plans daily chat and monthly payment caps for teenagers
ScatterLab plans default limits for teenage users of its Zeta AI character-chat service by year-end: three hours of use a day and 100,000 won in monthly payments. The service would prompt a break before the time limit and restrict further activity at the cap. The announcement follows scrutiny of harmful conversations that existing safeguards did not always stop.
Artificial Intelligence··Evening
Default limits are planned for teenagers
ScatterLab, the South Korean company operating the AI character-chat service Zeta, announced new controls for teenage accounts on October 12. It plans a default daily use limit of three hours and a monthly payment cap of 100,000 won by year-end. Near the time limit, the app would suggest taking a break. Once the relevant caps are reached, further use and payments would be restricted. These are planned controls rather than restrictions already in force.[1], [2]
Zeta lets users create characters and fictional worlds and talk with AI inside those settings. Kyunghyang Shinmun reports 1.48 million monthly active users at the end of September, including 380,000 aged 19 or younger. Those are platform audience figures, not counts of teens harmed or of people who will reach the new caps. The proposed limits address time and spending as separate parts of use.[1]
Existing safeguards faced a bypass test
The service already has a safe mode that restricts material considered harmful to teenagers. Its freer unlimited mode is available only when identity verification establishes adulthood. The interface continuously identifies the conversation partner as AI. Repeated expressions involving suicide or self-harm can trigger information on professional counselling organisations. ScatterLab is proposing the new time and payment caps alongside these existing content and identity controls.[1]
A check by Democratic Party lawmaker Hwang Jung-ah’s office examined safe-mode scenarios with underage characters. Some conversations involving sexual violence, exploitation and threats were generated. The harmful-conversation detector blocked roughly 200 problematic utterances during the check, but some harmful content still appeared during attempts to bypass restrictions. The reported test identifies gaps in the safeguards; it does not measure how often ordinary users encounter such content.[1]
Model and review changes are also promised
ScatterLab told the lawmaker’s office it would adjust generation from the story-creation stage so depictions better match the user’s age. It also plans to train against requests designed to bypass blocks and develop a detector that considers the wider conversation. Monthly adversarial checks are proposed. These measures target the content failures observed in the office’s test, while the separate default caps target length of use and payment.[1]
Chief executive Kim Jong-yoon says safe and healthy use by teenagers is a priority and that the company will continue talks with government and industry about user protection. ScatterLab says it reviewed approaches used by other digital services before setting its plan. The announced rollout remains due by the end of the year. No observed effect of the future caps on Zeta usage or harmful conversations is yet available from these accounts.[1], [2]
