Protecting youth as AI moves to action

From Sept 14 to 20, the 2026 China Cybersecurity Week took place across China. On Sept 14, the National Technical Committee 260 on Cybersecurity of Standardization Administration of China (SAC/TC260) released the AI Safety Governance Framework 3.0, following versions 1.0 in 2024 and 2.0 in 2025.

The latest version was updated in response to emerging trends in AI development and new challenges in safety governance. As AI evolves from simply providing information to interacting with users in increasingly varied and sophisticated ways, the online safety risks facing young people are changing as well.

One thing has become increasingly clear to me over the past year: AI's impact on young people is growing more complex and is evolving rapidly.

One important change is that potentially harmful AI-generated content can no longer always be identified through traditional methods.

In the past, the focus was largely on relatively obvious material involving violence, vulgarity or other harmful content in text and images.

Today, AI can draw on a child's conversations and interactions over time to generate highly personalized responses. These responses may contain no obvious sensitive words, yet they can subtly shape a young person's understanding of reality, responsibility and relationships.

Technologies such as AI-generated face swaps and synthetic videos can also be misused to deceive minors into revealing private or personal information, making online abuse and exploitation more difficult to detect.

Another development highlighted by the AI Safety Governance Framework 3.0 particularly concerns me: AI risks are shifting from content-related risks to behavior-related risks.

In the past, AI might "say the wrong thing" or generate an inaccurate image. Today, AI agents can use tools and carry out tasks across multiple systems.

For minors, this means AI is no longer simply telling them something — it may in fact do something on their behalf. It could, for example, send messages, initiate transactions or access sensitive data without the user fully understanding what is happening.

That is what makes the risks posed by AI agents particularly significant: actions taken in the virtual world can increasingly produce consequences in the real world.

For me, this points to another important issue: safeguards designed for minors depend in part on knowing who is, in actuality, using a service.

A registered account holder and the person using the account are not always the same. A young person may, for example, access a service through a parent's account or find other ways to enter an environment intended for adults.

When a system treats a minor as an adult, age-appropriate safeguards may not function as intended.

For this reason, I believe future regulation should continue to address harmful online content while also placing greater emphasis on identity, permissions and traceability.

This means developing proportionate mechanisms that can provide reasonable assurance about who is using a service, which AI system or agent was involved, what actions it carried out and whether those actions can be traced if something goes wrong.

From my experience providing legal services in the education sector, I have seen schools devote increasing attention to cybersecurity and risks associated with emerging technologies.

Cybersecurity education and legal-awareness programs are becoming a more regular part of school activities, while some schools have also begun offering AI-related courses.

Many parents, however, still have only a limited understanding of what their children are doing with AI or what kinds of online communities they are entering. Technology is changing so quickly that adults can sometimes struggle to keep up with new risks as quickly as young people adopt new tools.

For young people themselves, I have one simple piece of advice: if someone tells you, "Don't tell your parents," be on your guard.

Technology will continue to change, but open communication, protection of personal information and awareness of potential risks remain among the most fundamental safeguards for staying safe online.

Written by Lu Haitian, a partner at Beijing Zhanda Law Firm. He currently serves as legal counsel to the Dongcheng District Education Commission and the Dongcheng District Health Commission in Beijing, advising on legal matters on the front lines of youth education and public health.