Several leading AI companies are taking different approaches to reduce the risk that teenagers will harm themselves or others while using AI tools. Lauren Jonas, OpenAI’s head of youth well‑being, told Semafor the guiding principle is to “avoid the mistakes that were made before us.” Jonas also argued that “AI is not social media” and that teens mainly use AI tools for schoolwork.
Different principles and measures
OpenAI on Thursday published its position that teens should be able to access AI with built‑in safeguards, such as nudges to take breaks and parental time limits. OpenAI argues that allowing young people to practice with the technology in a safe way prepares them better for adult life.
Anthropic has taken a different route: it requires users to enter a birthdate indicating they are older than 18 to use its AI products. Meta announced that it will notify parents if their child discusses self‑harm with its chatbot, a step it said follows OpenAI’s lead.
Practical challenges of age verification
However, teenagers have little incentive to provide their real ages or to link their accounts to a parent’s profile, so requiring those steps alone does not fully prevent misuse. Several major AI companies have deployed machine‑learning systems that estimate users’ ages from their queries and flag accounts for additional verification — currently the most sophisticated technical approach to protecting minors.
Safety requires more than product changes
Experience from the past decade suggests youth safety depends on more than product updates. Effective protection requires support from communities, schools, parents and the young people themselves. Corporate measures — whether age gates, notifications, or ML‑based age estimates — can only go so far without complementary social and familial safeguards.
Summary
OpenAI, Anthropic and Meta are pursuing different strategies for protecting teenage users: OpenAI favors controlled access with behavioral safeguards, Anthropic enforces an 18+ age requirement, and Meta will notify parents about self‑harm conversations. Machine‑learning age estimation is increasingly used, but experts emphasize that technological fixes alone are insufficient to ensure teen safety.
