Empowering Parents, Protecting Teens: Meta's Approach to AI Safety

Key Findings

• — AI characters are designed to not engage in age-inappropriate discussion of self-harm, suicide, or disordered eating with teens, and are designed to respond safely and direct teens to expert resources when appropriate
• — Teen accounts are restricted to a limited set of AI characters focused on age-appropriate topics such as education, sports, and hobbies, excluding romantic or otherwise inappropriate content
• — New parental controls let parents turn off a teen's access to one-on-one AI-character chats entirely, block specific characters, and see topic-level insights into what teens discuss with AI characters and Meta's AI assistant
• — Meta uses AI-based age-prediction technology to apply teen protections even to users who claim to be adults but are suspected to be teens
• — As of the January 23, 2026 update, Meta paused teens' access to existing AI characters globally while developing a new version, with parental controls to apply once the new version ships

Methodology Notes

Company newsroom post with two dated in-place updates (rather than a separate follow-up document); no accompanying technical safety evaluation or benchmark is presented. The January 2026 update represents a materially stronger measure (a global pause) than the October 2025 original framing (parental opt-out controls).

Sources