Meta launches AI-controlled teen safety system with restricted mode

Meta rolled out a teen safety system in which its AI runs in a restricted mode by default for younger users — refusing sexual or romantic conversations, discussions of alcohol and harmful substances, and content promoting violence. New parental controls let parents monitor the topics their children discuss with the AI, adding oversight to an area of intense regulatory and public concern.

The move is defensive and reactive: AI companions and chatbots have drawn fierce criticism and legal scrutiny over interactions with minors, and xAI's ongoing child-exploitation class action this week underscores the industry's exposure. Meta, which operates AI across Instagram, WhatsApp, and Facebook — platforms with enormous teen user bases — faces particular pressure to demonstrate guardrails before regulators impose them.

Mechanically, the restricted mode is a policy-and-classifier layer that constrains the model's outputs for accounts identified as belonging to teens, paired with parental visibility into conversation topics (though notably topics, not full transcripts, preserving some teen privacy). The design tries to balance safety with the autonomy debate that surrounds monitoring minors' communications.

Competitively, teen safety is becoming table stakes — every consumer AI provider will need defensible guardrails, and Meta moving publicly sets a reference point. The skeptical read: restricted modes are only as good as their classifiers, and determined teens routinely circumvent filters; critics will test how easily the guardrails break. Watch for independent red-teaming of the restrictions, regulatory response, and whether other platforms match Meta's parental-monitoring features.