Legislation is taking shape that would impose specific protection obligations on generative AI chatbots that have a high share of teenage users, including age verification, parental consent and limits on time spent using the services.

The National Assembly’s Science, ICT, Broadcasting and Communications Committee has sent a related amendment to the Information Network Act to a bill review subcommittee. The Broadcasting, Media and Communications Committee also agrees with the bill’s intent and the need for specific safeguards, raising expectations that talks will accelerate on creating legal protections for teenage AI users, including AI character chats.

According to the National Assembly on Sept. 11, the committee referred on Sept. 8 to the bill review subcommittee a partial amendment to the Act on Promotion of Information and Communications Network Utilization and Information Protection, proposed by Democratic Party lawmaker In-cheol Cho (조인철).

The amendment would add generative AI-based services to the scope of the law, in addition to interactive information and communications services. It also defines specific measures to protect child and teenage users as operator obligations.

Under current law, operators are required to make efforts to prevent inappropriate information from being provided when offering interactive services to children under the age of 14.

The amendment would expand the protected group to all children and teenagers. Operators must verify a user’s age and identity, and if the user is a child or teenager, confirm the consent of a legal representative.

If a child, teenager or legal representative requests it, operators must limit how and when the service can be used. If the user requests it, operators must also provide details such as the time and content of conversations the user took part in.

If a child or teenager has been exposed, or is likely to be exposed, to risks such as suicide or self-harm, narcotics misuse or abuse, or sexual exploitation or abuse, the amendment would allow reports to be filed. Operators must also respond, including by taking steps to remove the risk, when a report is received.

Behind the bill is a view that AI chatbots have moved beyond simple information tools to play roles as teenagers’ friends, romantic partners and counselling counterparts, making it difficult to respond with existing protection rules alone.

Cho said in the bill’s statement of purpose that cases have occurred in which children and teenagers form emotionally dependent relationships with AI chatbots, or AI recommends suicide or self-harm and leads users into inappropriate sexual conversations. He said it is difficult to respond to risks such as encouraging suicide or self-harm, drug misuse or abuse, sexual exploitation and overdependence with only declaratory provisions in current law.

He said, "Generative AI has become a daily conversation partner and counselling channel for children and teenagers, but it is difficult to adequately protect them from dangerous conversations or harmful information with the current system alone." He added, "There is a need to establish practical safeguards so that from the service design stage, the developmental stage and safety of children and teenagers can be prioritised."

Cho’s office said it will actively push for passage of the bill during this regular session of the National Assembly. An official from his office said, "A bipartisan consensus has formed on the need to protect teenage AI users," and added, "Now that we have shared the bill’s intent with the floor leaders from both parties and others, we plan to actively discuss it in the subcommittee and continue the legislative process."

The Broadcasting, Media and Communications Committee also agrees with the bill’s intent and the need for specific protective measures.

An official at the committee said, "In the government review opinion submitted to the National Assembly after the bill was proposed, we stated that we strongly agree with the legislative intent, and we judged that the obligations specifically imposed by the bill are effective and necessary measures."

The official added, "Since the bill itself was prepared with AI character chat services in mind, relational chatbots and interactive chatbots are also included in the scope."

If passed, domestic AI character chats such as Zeta and Crack are expected to fall directly under the law. According to MobileIndex, based on monthly active users last month, users aged in their teens or younger accounted for 58.3 percent of Zeta and 47.8 percent of Crack, the No. 1 and No. 2 AI character chat services in South Korea.

Related operators are running their own safeguards, such as identity verification for adult content and youth protection policies. If the amendment passes, however, age and identity checks, legal representative consent, time limits and risk reporting and response measures that have so far been left to operator discretion would become legal obligations.

The amendment would take effect six months after its promulgation.