Frontier AI labs don’t want to be known for helping teens commit harm against themselves or others. “The principle here is to avoid the mistakes that were made before us,” Lauren Jonas, OpenAI’s head of youth well-being, told Semafor. “AI is not social media,” she said, arguing that teens primarily use its tools for schoolwork.
OpenAI on Thursday published its stance on why teens should have access to AI with safeguards like nudges to take breaks and time limits set by parents, saying kids will be less prepared for life as adults if they don’t practice with the technology when they are young. That’s a different approach than Anthropic’s, which requires users to enter a birthday that indicates they are more than 18 years old to use its AI products. Meta also just announced it will notify parents if their child discusses self-harm with its chatbot, following OpenAI’s lead.
Teens, however, need to buy into the idea by submitting their real ages and connecting a parent’s account — actions that they have little incentive to take. Major AI companies have employed machine learning that predicts users’ ages based on their queries, flagging accounts for additional verification, which is the most sophisticated method for protecting kids thus far. But if the last decade has shown us anything, it’s that teen safety is about more than product updates: It requires support from communities, schools, parents, and the kids themselves.