"As we continue to refine our systems, we're adding more guardrails as an extra precaution—including training our AIs not to engage with teens on these topics, but to guide them to expert resources, ...