Common Sense Media Flags Google AI Search as Unacceptable Risk After 2,600 Child-Safety Tests
Updated
Updated · PBS NewsHour · Jul 15
Common Sense Media Flags Google AI Search as Unacceptable Risk After 2,600 Child-Safety Tests
3 articles · Updated · PBS NewsHour · Jul 15
Summary
More than 2,600 test interactions led Common Sense Media to conclude Google's AI Overview and AI Mode pose an "unacceptable risk" to children, citing missed self-harm signals, unsafe advice, fabricated answers and homework completion.
The report found the built-in search tools violated seven of eight AI behavior principles and all five severe-harm red lines; AI Mode also completed all 180 assigned math and humanities tasks it was given.
Specific failures included AI Overview missing 29% of explicit suicide references, offering account-access steps after a goodbye-style query, and replying "grindset locked in!" to a child boasting of three days without sleep.
Google disputed the findings, saying the tests used narrow, contrived queries, that many highlighted responses could not be reproduced, and that Search includes guardrails plus parental controls such as SafeSearch and Family Link.
The warning carries extra weight because AI summaries are used by three-quarters of U.S. children ages 9 to 17 and Google's school tools power millions of classroom devices, while Congress and states weigh new AI child-safety rules.
If Google has safer AI technology in one mode, why isn't that safety standard the default for all children?
As AI becomes a child's 'confidant,' are we trading critical thinking and human connection for algorithmic convenience?
"Google Search AI Deemed 'Unacceptable Risk' for Kids: 2,600-Test Study Reveals Dangerous Flaws, Sparks Demands for Stronger Safeguards"
Overview
On July 15, 2026, the CSM Youth AI Safety Institute released a report declaring Google Search’s AI features, including AI Overview and AI Mode, an “unacceptable risk” for children. The report was based on over 2,600 test interactions using accounts set up for 11- and 15-year-olds with SafeSearch enabled. Despite these protections, the AI features generated dangerous and inappropriate responses to children’s queries. The report highlighted serious safety concerns, especially failures in mental health and crisis response, showing that Google’s AI did not provide safe guidance for sensitive topics. These findings raise urgent questions about the safety of AI in search for young users.