The Youth AI Safety Institute at Common Sense Media (California nonprofit) rated Alphabet’s Google Search AI features — AI Overviews and AI Mode — an “unacceptable risk” to children in a July 2026 risk assessment (report dated ~July 14–15; Japan Times coverage July 17).
Core findings:
- On test accounts configured for minors (11- and 15-year-old Family Link accounts with SafeSearch on), AI features failed to detect suicide risks, said an eating disorder symptom was normal, and provided instructions for creating deepfakes / sexually explicit fake content.
- Unlike standalone chatbots, Google’s AI tools are built into the default search experience on school and personal devices; administrators/parents currently have no way to disable them.
- AI Overviews generate automated answers above traditional results; AI Mode is a conversational search feature. Critics say the format presents AI summaries as definitive answers.
Additional detail from Common Sense / Axios / PBS coverage:
- Evaluation May 19–July 1, 2026; >2,600 test interactions; 2,100+ cited sources audited.
- Failed all five severe-harm “Red Lines”; Unacceptable/High Risk on seven of eight AI Principles.
- Crisis moments: AI Overview supplied appropriate resource in ~58% of cases; AI Mode ~77% — neither meets Common Sense’s safety standard.
- AI Mode completed 100% of homework assignments in tests.
- Features treated forums/social posts as comparable to medical institutions and peer-reviewed research; gave correct and incorrect answers with similar confidence.
- Google disputed findings via PBS response, calling tests “a narrow set of ambiguous and contrived queries that don’t reflect how people use Search.”
- Common Sense recommendation: give parents and schools control to disable/align AI search features; until then, digital literacy and alternative search when possible.